Strategies to Achieve Labor Flexibility in the Garment Industry

IJCSIIJCSI

International Journal of

Computer Science Issues

Volume 7, Issue 4, No 6, July 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814

© IJCSI PUBLICATION www.IJCSI.org

IJCSI proceedings are currently indexed by:

© IJCSI PUBLICATION 2010 www.IJCSI.org

IJCSI Publicity Board 2010 Dr. Borislav D Dimitrov Department of General Practice, Royal College of Surgeons in Ireland Dublin, Ireland Dr. Vishal Goyal Department of Computer Science, Punjabi University Patiala, India Mr. Nehinbe Joshua University of Essex Colchester, Essex, UK Mr. Vassilis Papataxiarhis Department of Informatics and Telecommunications National and Kapodistrian University of Athens, Athens, Greece

EDITORIAL In this fourth edition of 2010, we bring forward issues from various dynamic computer science areas ranging from system performance, computer vision, artificial intelligence, ontologies, software engineering, multimedia, pattern recognition, information retrieval, databases, security and networking among others. Considering the growing interest of academics worldwide to publish in IJCSI, we invite universities and institutions to partner with us to further encourage open-access publications. As always we thank all our reviewers for providing constructive comments on papers sent to them for review. This helps enormously in improving the quality of papers published in this issue. Apart from availability of the full-texts from the journal website, all published papers are deposited in open-access repositories to make access easier and ensure continuous availability of its proceedings. We are pleased to present IJCSI Volume 7, Issue 4, July 2010, split in nine numbers (IJCSI Vol. 7, Issue 4, No. 6). Out of the 179 paper submissions, 57 papers were retained for publication. The acceptance rate for this issue is 31.84%. We wish you a happy reading! IJCSI Editorial Board July 2010 Issue ISSN (Print): 1694-0814 ISSN (Online): 1694-0784 © IJCSI Publications www.IJCSI.org

IJCSI Editorial Board 2010 Dr Tristan Vanrullen Chief Editor LPL, Laboratoire Parole et Langage - CNRS - Aix en Provence, France LABRI, Laboratoire Bordelais de Recherche en Informatique - INRIA - Bordeaux, France LEEE, Laboratoire d'Esthétique et Expérimentations de l'Espace - Université d'Auvergne, France Dr Constantino Malagôn Associate Professor Nebrija University Spain Dr Lamia Fourati Chaari Associate Professor Multimedia and Informatics Higher Institute in SFAX Tunisia Dr Mokhtar Beldjehem Professor Sainte-Anne University Halifax, NS, Canada Dr Pascal Chatonnay Assistant Professor MaÎtre de Conférences Laboratoire d'Informatique de l'Université de Franche-Comté Université de Franche-Comté France Dr Karim Mohammed Rezaul Centre for Applied Internet Research (CAIR) Glyndwr University Wrexham, United Kingdom Dr Yee-Ming Chen Professor Department of Industrial Engineering and Management Yuan Ze University Taiwan

Dr Vishal Goyal Assistant Professor Department of Computer Science Punjabi University Patiala, India Dr Dalbir Singh Faculty of Information Science And Technology National University of Malaysia Malaysia Dr Natarajan Meghanathan Assistant Professor REU Program Director Department of Computer Science Jackson State University Jackson, USA Dr Deepak Laxmi Narasimha Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Kuala Lumpur, Malaysia Dr Navneet Agrawal Assistant Professor Department of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India Dr T. V. Prasad Professor Department of Computer Science and Engineering, Lingaya's University Faridabad, Haryana, India Prof N. Jaisankar Assistant Professor School of Computing Sciences, VIT University Vellore, Tamilnadu, India

IJCSI Reviewers Committee 2010 Mr. Markus Schatten, University of Zagreb, Faculty of Organization and Informatics, Croatia Mr. Vassilis Papataxiarhis, Department of Informatics and Telecommunications, National and Kapodistrian University of Athens, Athens, Greece Dr Modestos Stavrakis, University of the Aegean, Greece Dr Fadi KHALIL, LAAS -- CNRS Laboratory, France Dr Dimitar Trajanov, Faculty of Electrical Engineering and Information technologies, ss. Cyril and Methodius Univesity - Skopje, Macedonia Dr Jinping Yuan, College of Information System and Management,National Univ. of Defense Tech., China Dr Alexis Lazanas, Ministry of Education, Greece Dr Stavroula Mougiakakou, University of Bern, ARTORG Center for Biomedical Engineering Research, Switzerland Dr Cyril de Runz, CReSTIC-SIC, IUT de Reims, University of Reims, France Mr. Pramodkumar P. Gupta, Dept of Bioinformatics, Dr D Y Patil University, India Dr Alireza Fereidunian, School of ECE, University of Tehran, Iran Mr. Fred Viezens, Otto-Von-Guericke-University Magdeburg, Germany Dr. Richard G. Bush, Lawrence Technological University, United States Dr. Ola Osunkoya, Information Security Architect, USA Mr. Kotsokostas N.Antonios, TEI Piraeus, Hellas Prof Steven Totosy de Zepetnek, U of Halle-Wittenberg & Purdue U & National Sun Yat-sen U, Germany, USA, Taiwan Mr. M Arif Siddiqui, Najran University, Saudi Arabia Ms. Ilknur Icke, The Graduate Center, City University of New York, USA Prof Miroslav Baca, Faculty of Organization and Informatics, University of Zagreb, Croatia Dr. Elvia Ruiz Beltrán, Instituto Tecnológico de Aguascalientes, Mexico Mr. Moustafa Banbouk, Engineer du Telecom, UAE Mr. Kevin P. Monaghan, Wayne State University, Detroit, Michigan, USA Ms. Moira Stephens, University of Sydney, Australia Ms. Maryam Feily, National Advanced IPv6 Centre of Excellence (NAV6) , Universiti Sains Malaysia (USM), Malaysia Dr. Constantine YIALOURIS, Informatics Laboratory Agricultural University of Athens, Greece Mrs. Angeles Abella, U. de Montreal, Canada Dr. Patrizio Arrigo, CNR ISMAC, italy Mr. Anirban Mukhopadhyay, B.P.Poddar Institute of Management & Technology, India Mr. Dinesh Kumar, DAV Institute of Engineering & Technology, India Mr. Jorge L. Hernandez-Ardieta, INDRA SISTEMAS / University Carlos III of Madrid, Spain Mr. AliReza Shahrestani, University of Malaya (UM), National Advanced IPv6 Centre of Excellence (NAv6), Malaysia Mr. Blagoj Ristevski, Faculty of Administration and Information Systems Management - Bitola, Republic of Macedonia Mr. Mauricio Egidio Cantão, Department of Computer Science / University of São Paulo, Brazil Mr. Jules Ruis, Fractal Consultancy, The Netherlands

Mr. Mohammad Iftekhar Husain, University at Buffalo, USA Dr. Deepak Laxmi Narasimha, Department of Software Engineering, Faculty of Computer Science and Information Technology, University of Malaya, Malaysia Dr. Paola Di Maio, DMEM University of Strathclyde, UK Dr. Bhanu Pratap Singh, Institute of Instrumentation Engineering, Kurukshetra University Kurukshetra, India Mr. Sana Ullah, Inha University, South Korea Mr. Cornelis Pieter Pieters, Condast, The Netherlands Dr. Amogh Kavimandan, The MathWorks Inc., USA Dr. Zhinan Zhou, Samsung Telecommunications America, USA Mr. Alberto de Santos Sierra, Universidad Politécnica de Madrid, Spain Dr. Md. Atiqur Rahman Ahad, Department of Applied Physics, Electronics & Communication Engineering (APECE), University of Dhaka, Bangladesh Dr. Charalampos Bratsas, Lab of Medical Informatics, Medical Faculty, Aristotle University, Thessaloniki, Greece Ms. Alexia Dini Kounoudes, Cyprus University of Technology, Cyprus Mr. Anthony Gesase, University of Dar es salaam Computing Centre, Tanzania Dr. Jorge A. Ruiz-Vanoye, Universidad Juárez Autónoma de Tabasco, Mexico Dr. Alejandro Fuentes Penna, Universidad Popular Autónoma del Estado de Puebla, México Dr. Ocotlán Díaz-Parra, Universidad Juárez Autónoma de Tabasco, México Mrs. Nantia Iakovidou, Aristotle University of Thessaloniki, Greece Mr. Vinay Chopra, DAV Institute of Engineering & Technology, Jalandhar Ms. Carmen Lastres, Universidad Politécnica de Madrid - Centre for Smart Environments, Spain Dr. Sanja Lazarova-Molnar, United Arab Emirates University, UAE Mr. Srikrishna Nudurumati, Imaging & Printing Group R&D Hub, Hewlett-Packard, India Dr. Olivier Nocent, CReSTIC/SIC, University of Reims, France Mr. Burak Cizmeci, Isik University, Turkey Dr. Carlos Jaime Barrios Hernandez, LIG (Laboratory Of Informatics of Grenoble), France Mr. Md. Rabiul Islam, Rajshahi university of Engineering & Technology (RUET), Bangladesh Dr. LAKHOUA Mohamed Najeh, ISSAT - Laboratory of Analysis and Control of Systems, Tunisia Dr. Alessandro Lavacchi, Department of Chemistry - University of Firenze, Italy Mr. Mungwe, University of Oldenburg, Germany Mr. Somnath Tagore, Dr D Y Patil University, India Ms. Xueqin Wang, ATCS, USA Dr. Borislav D Dimitrov, Department of General Practice, Royal College of Surgeons in Ireland, Dublin, Ireland Dr. Fondjo Fotou Franklin, Langston University, USA Dr. Vishal Goyal, Department of Computer Science, Punjabi University, Patiala, India Mr. Thomas J. Clancy, ACM, United States Dr. Ahmed Nabih Zaki Rashed, Dr. in Electronic Engineering, Faculty of Electronic Engineering, menouf 32951, Electronics and Electrical Communication Engineering Department, Menoufia university, EGYPT, EGYPT Dr. Rushed Kanawati, LIPN, France Mr. Koteshwar Rao, K G Reddy College Of ENGG.&TECH,CHILKUR, RR DIST.,AP, India

Mr. M. Nagesh Kumar, Department of Electronics and Communication, J.S.S. research foundation, Mysore University, Mysore-6, India Dr. Ibrahim Noha, Grenoble Informatics Laboratory, France Mr. Muhammad Yasir Qadri, University of Essex, UK Mr. Annadurai .P, KMCPGS, Lawspet, Pondicherry, India, (Aff. Pondicherry Univeristy, India Mr. E Munivel , CEDTI (Govt. of India), India Dr. Chitra Ganesh Desai, University of Pune, India Mr. Syed, Analytical Services & Materials, Inc., USA Dr. Mashud Kabir, Department of Computer Science, University of Tuebingen, Germany Mrs. Payal N. Raj, Veer South Gujarat University, India Mrs. Priti Maheshwary, Maulana Azad National Institute of Technology, Bhopal, India Mr. Mahesh Goyani, S.P. University, India, India Mr. Vinay Verma, Defence Avionics Research Establishment, DRDO, India Dr. George A. Papakostas, Democritus University of Thrace, Greece Mr. Abhijit Sanjiv Kulkarni, DARE, DRDO, India Mr. Kavi Kumar Khedo, University of Mauritius, Mauritius Dr. B. Sivaselvan, Indian Institute of Information Technology, Design & Manufacturing, Kancheepuram, IIT Madras Campus, India Dr. Partha Pratim Bhattacharya, Greater Kolkata College of Engineering and Management, West Bengal University of Technology, India Mr. Manish Maheshwari, Makhanlal C University of Journalism & Communication, India Dr. Siddhartha Kumar Khaitan, Iowa State University, USA Dr. Mandhapati Raju, General Motors Inc, USA Dr. M.Iqbal Saripan, Universiti Putra Malaysia, Malaysia Mr. Ahmad Shukri Mohd Noor, University Malaysia Terengganu, Malaysia Mr. Selvakuberan K, TATA Consultancy Services, India Dr. Smita Rajpal, Institute of Technology and Management, Gurgaon, India Mr. Rakesh Kachroo, Tata Consultancy Services, India Mr. Raman Kumar, National Institute of Technology, Jalandhar, Punjab., India Mr. Nitesh Sureja, S.P.University, India Dr. M. Emre Celebi, Louisiana State University, Shreveport, USA Dr. Aung Kyaw Oo, Defence Services Academy, Myanmar Mr. Sanjay P. Patel, Sankalchand Patel College of Engineering, Visnagar, Gujarat, India Dr. Pascal Fallavollita, Queens University, Canada Mr. Jitendra Agrawal, Rajiv Gandhi Technological University, Bhopal, MP, India Mr. Ismael Rafael Ponce Medellín, Cenidet (Centro Nacional de Investigación y Desarrollo Tecnológico), Mexico Mr. Supheakmungkol SARIN, Waseda University, Japan Mr. Shoukat Ullah, Govt. Post Graduate College Bannu, Pakistan Dr. Vivian Augustine, Telecom Zimbabwe, Zimbabwe Mrs. Mutalli Vatila, Offshore Business Philipines, Philipines Dr. Emanuele Goldoni, University of Pavia, Dept. of Electronics, TLC & Networking Lab, Italy Mr. Pankaj Kumar, SAMA, India Dr. Himanshu Aggarwal, Punjabi University,Patiala, India Dr. Vauvert Guillaume, Europages, France

Prof Yee Ming Chen, Department of Industrial Engineering and Management, Yuan Ze University, Taiwan Dr. Constantino Malagón, Nebrija University, Spain Prof Kanwalvir Singh Dhindsa, B.B.S.B.Engg.College, Fatehgarh Sahib (Punjab), India Mr. Angkoon Phinyomark, Prince of Singkla University, Thailand Ms. Nital H. Mistry, Veer Narmad South Gujarat University, Surat, India Dr. M.R.Sumalatha, Anna University, India Mr. Somesh Kumar Dewangan, Disha Institute of Management and Technology, India Mr. Raman Maini, Punjabi University, Patiala(Punjab)-147002, India Dr. Abdelkader Outtagarts, Alcatel-Lucent Bell-Labs, France Prof Dr. Abdul Wahid, AKG Engg. College, Ghaziabad, India Mr. Prabu Mohandas, Anna University/Adhiyamaan College of Engineering, india Dr. Manish Kumar Jindal, Panjab University Regional Centre, Muktsar, India Prof Mydhili K Nair, M S Ramaiah Institute of Technnology, Bangalore, India Dr. C. Suresh Gnana Dhas, VelTech MultiTech Dr.Rangarajan Dr.Sagunthala Engineering College,Chennai,Tamilnadu, India Prof Akash Rajak, Krishna Institute of Engineering and Technology, Ghaziabad, India Mr. Ajay Kumar Shrivastava, Krishna Institute of Engineering & Technology, Ghaziabad, India Mr. Deo Prakash, SMVD University, Kakryal(J&K), India Dr. Vu Thanh Nguyen, University of Information Technology HoChiMinh City, VietNam Prof Deo Prakash, SMVD University (A Technical University open on I.I.T. Pattern) Kakryal (J&K), India Dr. Navneet Agrawal, Dept. of ECE, College of Technology & Engineering, MPUAT, Udaipur 313001 Rajasthan, India Mr. Sufal Das, Sikkim Manipal Institute of Technology, India Mr. Anil Kumar, Sikkim Manipal Institute of Technology, India Dr. B. Prasanalakshmi, King Saud University, Saudi Arabia. Dr. K D Verma, S.V. (P.G.) College, Aligarh, India Mr. Mohd Nazri Ismail, System and Networking Department, University of Kuala Lumpur (UniKL), Malaysia Dr. Nguyen Tuan Dang, University of Information Technology, Vietnam National University Ho Chi Minh city, Vietnam Dr. Abdul Aziz, University of Central Punjab, Pakistan Dr. P. Vasudeva Reddy, Andhra University, India Mrs. Savvas A. Chatzichristofis, Democritus University of Thrace, Greece Mr. Marcio Dorn, Federal University of Rio Grande do Sul - UFRGS Institute of Informatics, Brazil Mr. Luca Mazzola, University of Lugano, Switzerland Mr. Nadeem Mahmood, Department of Computer Science, University of Karachi, Pakistan Mr. Hafeez Ullah Amin, Kohat University of Science & Technology, Pakistan Dr. Professor Vikram Singh, Ch. Devi Lal University, Sirsa (Haryana), India Mr. M. Azath, Calicut/Mets School of Enginerring, India Dr. J. Hanumanthappa, DoS in CS, University of Mysore, India Dr. Shahanawaj Ahamad, Department of Computer Science, King Saud University, Saudi Arabia Dr. K. Duraiswamy, K. S. Rangasamy College of Technology, India Prof. Dr Mazlina Esa, Universiti Teknologi Malaysia, Malaysia

Dr. P. Vasant, Power Control Optimization (Global), Malaysia Dr. Taner Tuncer, Firat University, Turkey Dr. Norrozila Sulaiman, University Malaysia Pahang, Malaysia Prof. S K Gupta, BCET, Guradspur, India Dr. Latha Parameswaran, Amrita Vishwa Vidyapeetham, India Mr. M. Azath, Anna University, India Dr. P. Suresh Varma, Adikavi Nannaya University, India Prof. V. N. Kamalesh, JSS Academy of Technical Education, India Dr. D Gunaseelan, Ibri College of Technology, Oman Mr. Sanjay Kumar Anand, CDAC, India Mr. Akshat Verma, CDAC, India Mrs. Fazeela Tunnisa, Najran University, Kingdom of Saudi Arabia Mr. Hasan Asil, Islamic Azad University Tabriz Branch (Azarshahr), Iran Prof. Dr Sajal Kabiraj, Fr. C Rodrigues Institute of Management Studies (Affiliated to University of Mumbai, India), India Mr. Syed Fawad Mustafa, GAC Center, Shandong University, China Dr. Natarajan Meghanathan, Jackson State University, Jackson, MS, USA Prof. Selvakani Kandeeban, Francis Xavier Engineering College, India Mr. Tohid Sedghi, Urmia University, Iran Dr. S. Sasikumar, PSNA College of Engg and Tech, Dindigul, India Dr. Anupam Shukla, Indian Institute of Information Technology and Management Gwalior, India Mr. Rahul Kala, Indian Institute of Inforamtion Technology and Management Gwalior, India Dr. A V Nikolov, National University of Lesotho, Lesotho Mr. Kamal Sarkar, Department of Computer Science and Engineering, Jadavpur University, India Dr. Mokhled S. AlTarawneh, Computer Engineering Dept., Faculty of Engineering, Mutah University, Jordan, Jordan Prof. Sattar J Aboud, Iraqi Council of Representatives, Iraq-Baghdad Dr. Prasant Kumar Pattnaik, Department of CSE, KIST, India Dr. Mohammed Amoon, King Saud University, Saudi Arabia Dr. Tsvetanka Georgieva, Department of Information Technologies, St. Cyril and St. Methodius University of Veliko Tarnovo, Bulgaria Dr. Eva Volna, University of Ostrava, Czech Republic Mr. Ujjal Marjit, University of Kalyani, West-Bengal, India Dr. Prasant Kumar Pattnaik, KIST,Bhubaneswar,India, India Dr. Guezouri Mustapha, Department of Electronics, Faculty of Electrical Engineering, University of Science and Technology (USTO), Oran, Algeria Mr. Maniyar Shiraz Ahmed, Najran University, Najran, Saudi Arabia Dr. Sreedhar Reddy, JNTU, SSIETW, Hyderabad, India Mr. Bala Dhandayuthapani Veerasamy, Mekelle University, Ethiopa Mr. Arash Habibi Lashkari, University of Malaya (UM), Malaysia Mr. Rajesh Prasad, LDC Institute of Technical Studies, Allahabad, India Ms. Habib Izadkhah, Tabriz University, Iran Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University Bhilai, India Mr. Kuldeep Yadav, IIIT Delhi, India Dr. Naoufel Kraiem, Institut Superieur d'Informatique, Tunisia

Prof. Frank Ortmeier, Otto-von-Guericke-Universitaet Magdeburg, Germany Mr. Ashraf Aljammal, USM, Malaysia Mrs. Amandeep Kaur, Department of Computer Science, Punjabi University, Patiala, Punjab, India Mr. Babak Basharirad, University Technology of Malaysia, Malaysia Mr. Avinash singh, Kiet Ghaziabad, India Dr. Miguel Vargas-Lombardo, Technological University of Panama, Panama Dr. Tuncay Sevindik, Firat University, Turkey Ms. Pavai Kandavelu, Anna University Chennai, India Mr. Ravish Khichar, Global Institute of Technology, India Mr Aos Alaa Zaidan Ansaef, Multimedia University, Cyberjaya, Malaysia Dr. Awadhesh Kumar Sharma, Dept. of CSE, MMM Engg College, Gorakhpur-273010, UP, India Mr. Qasim Siddique, FUIEMS, Pakistan Dr. Le Hoang Thai, University of Science, Vietnam National University - Ho Chi Minh City, Vietnam Dr. Saravanan C, NIT, Durgapur, India Dr. Vijay Kumar Mago, DAV College, Jalandhar, India Dr. Do Van Nhon, University of Information Technology, Vietnam Mr. Georgios Kioumourtzis, University of Patras, Greece Mr. Amol D.Potgantwar, SITRC Nasik, India Mr. Lesedi Melton Masisi, Council for Scientific and Industrial Research, South Africa Dr. Karthik.S, Department of Computer Science & Engineering, SNS College of Technology, India Mr. Nafiz Imtiaz Bin Hamid, Department of Electrical and Electronic Engineering, Islamic University of Technology (IUT), Bangladesh Mr. Muhammad Imran Khan, Universiti Teknologi PETRONAS, Malaysia Dr. Abdul Kareem M. Radhi, Information Engineering - Nahrin University, Iraq Dr. Mohd Nazri Ismail, University of Kuala Lumpur, Malaysia Dr. Manuj Darbari, BBDNITM, Institute of Technology, A-649, Indira Nagar, Lucknow 226016, India Ms. Izerrouken, INP-IRIT, France Mr. Nitin Ashokrao Naik, Dept. of Computer Science, Yeshwant Mahavidyalaya, Nanded, India Mr. Nikhil Raj, National Institute of Technology, Kurukshetra, India Prof. Maher Ben Jemaa, National School of Engineers of Sfax, Tunisia Prof. Rajeshwar Singh, BRCM College of Engineering and Technology, Bahal Bhiwani, Haryana, India Mr. Gaurav Kumar, Department of Computer Applications, Chitkara Institute of Engineering and Technology, Rajpura, Punjab, India Mr. Ajeet Kumar Pandey, Indian Institute of Technology, Kharagpur, India Mr. Rajiv Phougat, IBM Corporation, USA Mrs. Aysha V, College of Applied Science Pattuvam affiliated with Kannur University, India Dr. Debotosh Bhattacharjee, Department of Computer Science and Engineering, Jadavpur University, Kolkata-700032, India Dr. Neelam Srivastava, Institute of engineering & Technology, Lucknow, India Prof. Sweta Verma, Galgotia's College of Engineering & Technology, Greater Noida, India Mr. Harminder Singh BIndra, MIMIT, INDIA Dr. Lokesh Kumar Sharma, Chhattisgarh Swami Vivekanand Technical University, Bhilai, India Mr. Tarun Kumar, U.P. Technical University/Radha Govinend Engg. College, India Mr. Tirthraj Rai, Jawahar Lal Nehru University, New Delhi, India

Mr. Akhilesh Tiwari, Madhav Institute of Technology & Science, India Mr. Dakshina Ranjan Kisku, Dr. B. C. Roy Engineering College, WBUT, India Ms. Anu Suneja, Maharshi Markandeshwar University, Mullana, Haryana, India Mr. Munish Kumar Jindal, Punjabi University Regional Centre, Jaito (Faridkot), India Dr. Ashraf Bany Mohammed, Management Information Systems Department, Faculty of Administrative and Financial Sciences, Petra University, Jordan Mrs. Jyoti Jain, R.G.P.V. Bhopal, India Dr. Lamia Chaari, SFAX University, Tunisia Mr. Akhter Raza Syed, Department of Computer Science, University of Karachi, Pakistan Prof. Khubaib Ahmed Qureshi, Information Technology Department, HIMS, Hamdard University, Pakistan Prof. Boubker Sbihi, Ecole des Sciences de L'Information, Morocco Dr. S. M. Riazul Islam, Inha University, South Korea Prof. Lokhande S.N., S.R.T.M.University, Nanded (MH), India Dr. Vijay H Mankar, Dept. of Electronics, Govt. Polytechnic, Nagpur, India Dr. M. Sreedhar Reddy, JNTU, Hyderabad, SSIETW, India Mr. Ojesanmi Olusegun, Ajayi Crowther University, Oyo, Nigeria Ms. Mamta Juneja, RBIEBT, PTU, India Dr. Ekta Walia Bhullar, Maharishi Markandeshwar University, Mullana Ambala (Haryana), India Prof. Chandra Mohan, John Bosco Engineering College, India Mr. Nitin A. Naik, Yeshwant Mahavidyalaya, Nanded, India Mr. Sunil Kashibarao Nayak, Bahirji Smarak Mahavidyalaya, Basmathnagar Dist-Hingoli., India Prof. Rakesh.L, Vijetha Institute of Technology, Bangalore, India Mr B. M. Patil, Indian Institute of Technology, Roorkee, Uttarakhand, India Mr. Thipendra Pal Singh, Sharda University, K.P. III, Greater Noida, Uttar Pradesh, India Prof. Chandra Mohan, John Bosco Engg College, India Mr. Hadi Saboohi, University of Malaya - Faculty of Computer Science and Information Technology, Malaysia Dr. R. Baskaran, Anna University, India Dr. Wichian Sittiprapaporn, Mahasarakham University College of Music, Thailand Mr. Lai Khin Wee, Universiti Teknologi Malaysia, Malaysia Dr. Kamaljit I. Lakhtaria, Atmiya Institute of Technology, India Mrs. Inderpreet Kaur, PTU, Jalandhar, India Mr. Iqbaldeep Kaur, PTU / RBIEBT, India Mrs. Vasudha Bahl, Maharaja Agrasen Institute of Technology, Delhi, India Prof. Vinay Uttamrao Kale, P.R.M. Institute of Technology & Research, Badnera, Amravati, Maharashtra, India Mr. Suhas J Manangi, Microsoft, India Ms. Anna Kuzio, Adam Mickiewicz University, School of English, Poland Dr. Debojyoti Mitra, Sir Padampat Singhania University, India Prof. Rachit Garg, Department of Computer Science, L K College, India Mrs. Manjula K A, Kannur University, India Mr. Rakesh Kumar, Indian Institute of Technology Roorkee, India

TABLE OF CONTENTS 1. QoS Assurance for Service-Based Applications Using Discrete-Event Simulation Yassine Jamoussi, Maha Driss, Jean-Marc Jézéquel and Henda Hajjami Ben Ghézala 2. Sentence Recognition Using Hopfield Neural Network Bipul Pandey, Sushil Ranjan, Anupam Shukla and Ritu Tiwari 3. Review Strategies and Analysis of Mobile Ad Hoc Network- Internet Integration Solutions Rakesh Kumar, Anil K. Sarje and Manoj Misra 4. A New Approach to Supervise Security in Social Network through Quantum Cryptography and Non-Linear Dimension Reduction Techniques Lokesh Jain and Satbir Jain 5. The Grassmannian Manifold and Controllability of the Linear Time-Invariant Control Systems S. M. Deshmukh, Seema S. Deshmukh, R. D. Kanphade and N. A. Patil 6. Strategies to Achieve Labor Flexibility in the Garment Industry Parisima Nassirnia and Masine Md. Tap

Pg 1-11 Pg 12-17 Pg 18-28 Pg 29-35 Pg 36-40 Pg 41-48

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 4, No 6, July 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814

1

QoS Assurance for Service-Based Applications

Using Discrete-Event Simulation

Yassine Jamoussi1, Maha Driss1,2, Jean-Marc Jézéquel2, and Henda Hajjami Ben Ghézala1

1ENSI, RIADI-GDL Laboratory, University of Manouba La Manouba, 2010, Tunisia

2IRISA/INRIA, University of Rennes I Rennes, 35042, France

Abstract

The new paradigm for distributed computing over the Internet is that of Web services. The goal of Web services is to achieve universal interoperability between applications by using standardized protocols and langua ges. One of the key ideas of the Web service paradigm is the ability of building com plex and value-added service-based applications by composing pre-existing services. For a s ervice-based application, in addition to its functional requirements, Quality of service (QoS) requirements are important and de serve a s pecial attention. In this paper, we introduce a discrete-event modeling approach for service-based application. This approach is oriented towards QoS assurance through discrete-event simulation. Keywords: Web Services, Service-based applications, QoS assurance, Discrete-event simulation.

1. Introduction

In th e last ten years, th e Service-Oriented Architecture (SOA) emerged as a powerful solution to enable interoperability between dist ributed software com ponents known as Web Servi ces (W Ss) [1, 2] . W Ss are universally accessible softwa re components that are advertised, discovered, and i nvoked t hrough t he W eb. The key aspect of the SOA is the use of standard technologies such as: WSDL, UDDI, and SOAP. These technologies define standard ways of W Ss definition, discovery, and invocation. SOA is th e b est so lution fo r composite application integration. Indeed, W Ss m ay be easi ly composed/aggregated together in to a n ew ap plication, regardless specific im plementation pl atforms and technologies [3]. The obtained Service-Based Application (SBA) may be further published as a new service creating a col laboration network between different organizations. For exam ple, telecom munication com panies can be

considered as an example of service aggregators [4]. Multiple an d d ifferent serv ices su ch as calling services (e.g., call forwardi ng and cal l barri ng), m essaging services (e.g., t ext messaging and vi deo messaging), and internet servi ces (e.g., chat and e-m ail) are brought together and offered via telephone. For an SBA, i n addi tion t o i ts funct ional requi rements, Quality of Service (QoS) requirements are im portant and deserve a speci al at tention. QoS requi rements for SB As include response tim e, throughput, availability and security [4, 5]. Being able to characterize SBAs based on QoS has three distinct advantages [6]:

It allows for the design of SBAs according to QoS requi rements. Indeed, i t i s important for service provi ders t o know t he QoS of a SBA at prior before offering it to their clients.

It allows for the selection and the execution of SBAs based on t heir QoS. Si nce m any servi ces provide o verlapping o r id entical functionality, different SB As can be com posed, sat isfying t he same funct ional requi rement. A choi ce needs to be made to determine which SBA is to be used to provide with the more beneficial QoS.

It allo ws fo r th e ev aluation o f altern ative adaptation strategies. The dy namic and unpredictable nat ure of the execution environment (e.g., network resources and devices characteristics) has an im portant im pact on QoS of SBAs. Th us, in o rder to b etter fu lfill QoS requirements, it is n ecessary to adapt SBAs in response t o an unexpect ed evol ution of the execution environment.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 4, No 6, July 2010 www.IJCSI.org

2

To assure the desired QoS requirements for an SBA, different analytical q uality assu rance tech niques can b e used. The goal of these techniques is to evaluate QoS and uncover quality defects in the SBAs after th ey have been created. An ex ample fo r an alytical q uality assu rance techniques i s si mulation. The goal of the simulation technique is to emulate the conversational behavior of the atomic WSs of an SB A. In this paper, we adopt a special case of simulation that is Discrete-Event Simulation (DES) to assure QoS of SBAs. DES has proved i ts effectiveness for diagnosing the QoS of software applications [7, 8]. To perform DES, we propose a di screte-event modeling approach for SBAs. This approach enables analytical description of SB As and al lows QoS evaluation in different status and condi tions of t he execut ion environment. To assure QoS eval uation, we defi ne a lightweight quality model fo r WSs focusing on essen tial properties of QoS that play a critical role for the effective management of WSs. These propert ies are m easured by DES technique. W e propose al so a cont ext m odel t hat supports an expl icit descri ption of t he execution environment. Th is m odel is d epicted in to th e sim ulation model i n order t o provi de a cont ext-based approach for evaluating QoS of SBAs. Our approach i s supported by a simulation framework named SBAS. The remainder of the paper i s st ructured as fol lows: Section 2 i ntroduces an overvi ew of the Web services architecture. Section 3 revi ews QoS assurance techniques for SBAs. In sect ion 4, di screte-event si mulation i ssues are addressed. W e descri be, i n Sect ion 5, t he proposed context model. In Section 6 , we in troduce o ur q uality model and explain metrics used t o m easure consi dered QoS properties. Section 7 i ntroduces our si mulation framework SBAS. Case study experi mental resul ts are documented in Section 8. Thi s paper ends wi th concluding remarks and future work.

2. Web Services Architecture Overview

SOA is an architecture that functions are defined as WSs. According to [1, 2], W Ss are sel f-contained, m odular applications that can be descri bed, published, located, and invoked over a net work, generally, the World Wide Web. The SOA i s descri bed t hrough t hree di fferent rol es: service provider, service request er and servi ce regi stry. SOA requi res t hree fundam ental operat ions: publishing, finding, and binding. The key i dea of t he SOA i s t he following: A servi ce provi der publ ishes servi ces i n a service registry . The servi ce requester searches for a service in the registry. He fi nds one or more by browsing or querying the registry. The servi ce request er uses t he service description to bind a service. These ideas are shown in the following figure 1. The above operat ions are

supported by st andard t echnologies that are: UDDI, WSDL, and SOAP [2].

Universal Description, Discovery, and Integration (UDDI) [9] : provides a registry where service providers can register and publish their services.

Web Servi ces Descri ption Language (WSDL) [10]: is an XM L based l anguage for descri bing WSs. It specifies the location of the W S and the operations exposed by the WS.

Simple Object Access Prot ocol (SOAP) [11] : i s an XML based prot ocol for exchangi ng information between WSs or between a client and a W S in a decentralized and distributed environment.

Fig. 1 Service-Oriented Architecture

What makes the SOA attractiv e is th e ab ility o f creatin g SBAs by composing existing WSs. Such a com position is based on the com mon standards of W S interfaces regardless of the languages that are used t o implement the WSs and the platform s where the WSs are executed. In general, t he W Ss have t he fol lowing feat ures t hat make them bet ter i n com position inside the heterogeneous environments [3]:

Loosely coupled: WSs are aut onomous and can operate i ndependently one from anot her. The loosely coupl ed feat ure enabl es W Ss to locate and communicate with each other dynamically at runtime.

Universal accessibility: W Ss can be defined, described and di scovered t hrough the Web that enables an easy accessibility.

Standard l anguages: W Ss are descri bed by standard XM L l anguages t hat have been considered as parts of the Web technology.


3

3. Quality of Service Assurance techniques for Service-Based Applications

By QoS, we refer t o non-functional requirements of SBAs such as response tim e, throughput, availability, and security [4, 5] . Thanks t o t he dynamic and unpredictable nature of the execution environment, the management and assurance of the QoS aspects of SBAs become of utmost importance. To achi eve t he desi red QoS of an SB A, two complementary kinds of t echniques can be em ployed: constructive and analytical q uality assu rance tech niques. Figure 2 provides an overview of these techniques.

Fig. 2 Overview of Quality Assurance Techniques for Service-Based Applications

The goal of constructive quality assurance techniques is to ensure QoS and prevent the introduction of quality defects while the SBA is created. Exam ples of such techniques include code generation, software development guidelines, as well as tem plates. Th e g oal o f an alytical q uality assurance t echniques i s t o eval uate QoS and uncover quality defects in an SBA after it h as b een created . W e sub-divide the analytical quality assurance techniques into three major classes: static analysis, monitoring, and testing. These classes have been proposed in the software quality assurance literature [12, 13] an d h ave b een u sed in a recent overview of quality assurance approaches for SBAs [14]. This section will co ver th e state o f th e art in an alytical quality assurance techniques for SBAs as our work deals with proposing DES as a new anal ytical testing technique to assure QoS for SBAs.

3.1 Static Analysis

Numerous effort s have been m ade by leading research groups to use static anal ysis t o eval uate QoS and t o uncover quality defects in SBAs.

The aim of static analysis is to systematically examine an SBA in o rder to ascertain wh ether some predefined QoS properties are m et. Examples of static analysis techniques include form al ones, l ike dat a fl ow anal ysis, m odel checking, sym bolic execu tion, t ype checki ng and correctness proofs. W e present , i n the following paragraphs, som e rel evant approaches appl ying st atic analysis to SBAs. In [15], Nakajima uses the model-checker SPIN to verify a set of QoS propert ies rel ated t o SBAs. SPIN provides a specification l anguage t hat descri bes t he SB A t o be a collection o f au tomata. Th e p roperties to be checked are reachability, deadlock, and freedom. These QoS properties are expressed as formulas of linear temporal logic. In [16] , Salaün et al. propose an approach t hat uses process algebra as an abstr act representation m eans to describe, compose, and reason on SB As. The t echniques used to check whether an SB A descri bed i n process-algebraic notations respects temporal QoS propert ies (e.g. safety and liveness) are referred to as model checking methods. In [17], Foster et al. propose the tool LTSA-WS to verify SBAs. This tool support s veri fication of QoS propert ies (e.g. absence of deadlock and liveness) created from design specifications and implementation models of SBAs to confirm expect ed QoS resul ts from t he vi ewpoints of both the designer and implementer. Scenarios are modeled in UML, in the form of message sequence charts, and then compiled into the finite state process algebra to concisely model t he requi red choreography behavi or and t o veri fy the required QoS properties. In [18, 19] , Kazhamiakin et al. and Osterweil ad dress the problem of t he veri fication and t he analysis of SBAs defined as a set of behavioral models against various QoS requirements. The works focus on modeling and analyzing specific QoS behavi oral propert ies of SB As, namely asynchronous com munications, dat a and t ime-related properties. In [18] , Kazhamiakin et al. present a framework which relies on a formal model where temporal logics are expl oited for the specification and the verification of the above QoS behavioral properties.

3.1 Monitoring

Monitoring has been widely used in many disciplines and, in particular, in service-oriented engineering. Monitoring is defined as a process of observing, collecting, and reporting information about t he execut ion and t he evolution of SBAs. The rel evant references of m onitoring are summarized in the following paragraphs. In [20], Keller and Ludwig propose the WSLA framework for the specification and t he m onitoring of servi ce-level agreements. The WSLA framework defines a language for


4

the specification of contract information th at allo ws fo r describing t he part ies i nvolved i n t he agreem ent, t he relevant QoS propert ies, as wel l as t he way s t o observe and m easure t hem and t he obl igations and constraints imposed on these properties. In [21] , Ludwi g et al. propose an archi tecture and implementation for the creation, the management, and t he monitoring of servi ce-level agreements represented as WS-Agreement documents. W S-Agreement specification provides a standardized way of defi ning cont ractual information bet ween servi ce provi der and cust omer. The proposed architecture i s cal led CREMONA. The monitoring m odule of CREMONA i s not onl y used to observe and detect contract violations, but also to predict future vi olations and t o engage appropriate adaptation strategies in advance. In [22] , Curbera et al. propose t he COLOMBO platform for devel oping, depl oying, and execut ing SB As. The COLOMBO platform in corporates th e to ols an d facilities for checki ng, m onitoring, and enforci ng servi ce requirements expressed i n W S-Policy not ations. W S-Policy notations define t he QoS assert ions t hat can be attached to a particular WS, operation, or a message type. In [23] , B aresi and Gui nea propose the run-time monitoring fram ework DYNAMO. DYNAMO uses an expressive monitoring language nam ely W SCoL for specifying m onitoring rules. DYNAMO oversees the execution of SB As by checki ng m onitoring rul es and by reacting as soon as t hey are vi olated by m eans of t he associated adaptation strategies. In [24] , B aresi et al. extend this work for what concerns t he kind of properties the approach can m onitor. The extended specification language, nam ely Ti med W SCoL, al lows for specifying temporal QoS propert ies over the events that occur during the SBA execution.

3.1 Testing

Testing is a frequently used technique for the analysis and the prediction of QoS of SBAs. The g oal o f testin g is to system atically ex ecute SBAs in order t o uncover QoS defect s. During testing, the SBA which i s t ested i s fed wi th concret e inputs and the produced outputs are observed. The observed out puts can deviate from the expected outputs wi th respect to functionality as well as Qo S. When the observed outputs deviate from the expected outputs, a defect is uncovered. A special case of testing is sim ulation. Simulation allows us to predict so ftware ap plications p erformance in different status and l oad condi tions of t he execut ion environment. The predicted results are used t o provi de feedback on the efficiency of th e ap plication. Simulating SBAs fo r Qo S ev aluation is a research area with little

previous work. Works in simulation that are the closest to ours are described by [25], [26], and [27]. In [25], Narayanan and M cIlraith propose a m odel-theoretic sem antics as well as distributed operational semantics th at can b e u sed for the simulation, the validation, the verification, the automated composition and the enactment of DAMLS-described SBAs. To provide a full service description, Narayanan and M cIlraith use the machinery o f situ ation calculus and its execution behaviour descri bed wi th Pet ri Net s. They use the simulation and m odeling envi ronment KarmaSIM to translate DAML-S markups to situation calculus and Petri Nets. In th is wo rk, th ree QoS properties are analyzed: reachability, liveness and the existence of deadlocks. In [26], Chandrasekaran et al. focus on problems related to SBA speci fication, eval uation, and execut ion usi ng Service Composition and Execution Tool (SCET). SCET allows to co mpose statically a W S p rocess with WSFL and t o generat e a si mulation m odel t hat can be processed by t he JSIM sim ulation en vironment. In this work, Chandrasekaran et al. have enhanced W SFL t o include QoS measures obtained by performing simulation tests. In [27], Mancini et al. present a framework which is aimed at support ing t he devel opment of self-optimizing, predictive and autonomic systems for W S architectures. It adopts a si mulation-based m ethodology whi ch al lows predicting QoS propert ies i n di fferent st atus and l oad conditions. In cont rast t o [25] and [26] , this work considers execut ion envi ronment i nformation i n t he simulation models. This work focuses on si mulating only atomic WSs. It proposes al so onl y one possi ble QoS optimization that is response t ime m inimization. Enhancements are needed to sim ulate SBAs and to add more optimization rules for QoS properties.

4. Discrete-Event Simulation Modeling of Service-Based Applications

Our work deal s wi th usi ng simulation as an analytical testing technique to assure QoS of SBAs. There are t wo m ain reasons for adopt ing si mulation techniques: fi rst, si mulation i s a dy namic anal ytical technique that al lows QoS predi ctions for soft ware applications in different st atus and condi tions of t he execution environment. Second, si mulation allows to tune and to evaluate software applications without experiencing the co st o f en acting th em. Th e o riginality o f o ur wo rk is the adoption of a special case of simulation that is the Discrete-Event Si mulation (DES) t o evaluate and assure QoS of SBAs. In this paper, we propose a SBA modeling approach that is oriented towards QoS eval uation through DES. DES i s a


5

kind of qualitative description o f a d ynamic system th e behavior of whi ch i s event -driven. Thi s t echnique i s frequently used to analyze and predict the QoS of software applications. Gi ving t he evol ution of the operation of an application, we can analyze its behavi or and eval uate appropriate quality measures. [7] and [8] are fundam ental works about discrete-event sy stems di agnosis. DES i s suitable t o m odel t he behavi or of a SB A si nce it is composed of W Ss which are decentralized and dynam ic. The i nteractions bet ween W Ss can be m odeled by a synchronized composition of several local models. To elaborate our simulation model for SBAs, we are based on the work present ed i n [28] t hat focuses on m odeling distributed applications. We model an SB A as a combination o f two typ es o f en tities: distributed application and n etwork in frastructure en tities [2 9]. Ou r simulation model is shown in figure 3.

Fig. 3 Discrete-Event Simulation Model for Service-Based Applications

4.1 Distributed Application Modeling

The operation of the distributed applications is based on the client-server model. In th is model, the client sends a set o f requests to the server and t he server sends a response back to the client for each request. The operation scenario is supported through specifying groups of actions: Processing: indicating data processing; Request: indicating invocation of a server process; Write: indicating data storage; Read: indicating data retrieval; Transfer: indicating data transfer between client and

server processes; Synchronize: indicating replica synchronization.

Each W S i s execut ed on a processing node. Processing action i ndicates i nvocation of t he processi ng uni t of the corresponding node and i s charact erized by t he am ount of data to be processed. Request action indicates invocation of a server process and is characterized by the nam e of the server, the nam e of the WS, its invoked interface and th e required inputs. Request action implies activation of t he net work, si nce t he request and the reply must be t ransferred from the invoking to the invoked process, and vice versa. There are two available actions for data storing, reading and writing, which are respectively characterized by the amount of the stored and retrieved data and the invoked server. The observations and perform ance anal ysis of SB As have proven that SOAP messages are small and simple [30]. A transfer action is used to indicate SOAP messages exchanged between processes. A synchronize action is needed since the replication of data is a com mon practice in such di stributed appl ications. Synchronize act ion parameters include the process replicas that m ust be sy nchronized and t he am ount of t ransferred data. To describe the operation of a SBA, we proceed by transforming the process behavior written in BPEL [31] into discrete-event actions. BPEL is a standard proposed by IBM and Microsoft along with several other companies to model composed Web services. BPEL defi nes a gram mar for describing the behavior of a SB A. It is composed of fifteen activity typ es, so me o f th em are b asic activ ities an d th e others are structured activitie s. Among the basic activities, the most important ones are the following: The <receive> activity: is for accepting the triggering

message from another WS; The <reply> activity: is for returning the response to

its requestor; The <invoke> activity: is for invoking another WS.

The structured activities define the execution orders of the activities inside their scopes. For example: The <sequence> act ivity: defines the sequential order

of the activities inside its scope; The <flow> activity: d efines th e co ncurrent relatio ns

of the activities inside its scope.

Each activity can be translated in to th e d iscrete-event formalism as one or several actions. Basic activities involve pro cessing, request, and data storing actions, while structured ones involve transferered and synchronized actions.


6

4.1 Network Infrastructure Modeling

In t he proposed m odeling scheme, the network infrastructure i s consi dered as a collection of individual networks and i nternetworks, exchangi ng m essages through relay nodes (act ive com munication devi ces e.g. routers and swi tches). Communication channels represent protocol suites (i.e. routing protocols (OSI layers 2 and 3) and peer-to-peer protocols (OSI layers 4-7)). According to the SOA, com munication between WSs is performed through exchanging SOAP messages. figure 4 illustrates one way o f making a remote call u sing SOAP in OSI network reference model [32].

Fig. 4 Sending a SOAP Request under OSI

First at application level, a native data object needs to be serialized into XML as SOAP request. Then, the SOAP message is passed to HTTP level. The HTTP layer, o n the client-side, needs to “hands hake” wi th servi ce-side by sending a “POST” req uest. Th is req uest initiated a TCP connection. Once receiving “HTTP: ACK”, the client-side HTTP begi ns t o send t he whol e SOAP m essage via TCP/IP. The SOAP message may be partitioned into a set of small segments at TCP layer. Appropriate headers and footers are attached to each segment as the segm ents are passed t hrough Transport , Net work, Dat a Li nk l ayers, until reaching the Network In terface Card (NIC) at the physical layer. The NIC is responsible for putting the packages onto the wire at a specific speed (network bandwidth) to next network device (such as a rout er or a switch), till serv er NIC [3 2]. The path from bottom (physical layer) to th e to p (ap plication layer) o n th e service-side i s opposi te t o t he process on the client-side: the received packages are unpacked at each layer and forwarded to next layer for further retrieving.

4.1 Context Model

By the term “context”, we mean “information utilized by the web servi ce t o adjust execut ion and out put to provide the client with a customized and personalized behavior”[33]. Since SB As are operat ing i n dy namic environments, variations of execut ion cont ext l ead t o vari ations in QoS expectations. In t his work, we propose a context-based approach for evaluating SBAs perform ances. We consider a context model which consists of a set of el ements grouped in 2 axes.

Fig. 5 Context Model

User context: describes user preferences. To each QoS property, user at tributes a wei ght. He chooses the value of this weight according to the level of the QoS propert y he needs. (e.g., execut ion time0.8).

Computing context: describes network connectivity (e.g., Internet connectivity, locality, and bandwidth) and devices capabilities (e.g., memory capacity and CPU speed).

Context information are described in the simulation model. Context changes are modeled as discrete events.

4.2 QoS Model

In [34], we d efine a lig ht-weight quality model focusing on essential p roperties o f Qo S th at p lay critical role for the effective management of W Ss and t hat can be m easured by DES t echnique. The QoS propert ies det ailed above are defined in the context of atomic WSs. They are also used to evaluate the QoS of composite WSs. To provide aggregation functions for computing the QoS of composite WSs, we use the QoS com putation models described by [6] and [35]. In these works, authors propose aggregation formulae for each pair QoS property/control statement (e.g., Sequence, Switch, Flow, and Loop). QoS aggregation functions are summarized in table 1.

Response Time: it corresponds t o t he t otal t ime needed by a W S to transform a set of inputs into


7

outputs. Response Time (RT) for a service s can be computed as follows:

RT(s) = ST(s) + DT(s) (1)

– Service Time (ST) is the time that the WS takes to perform its task.

– Delay Tim e ( DT) is th e tim e taken to send/receive SOAP messages.

Reliability: it corresponds to the likelihood that the service will p erform for its u sers on demand. Reliability ( R) of a service s i s funct ion of t he Failure Rate (FR):

R(s) = (1-FR(s))*100 (2) – FR= successful executions/scheduled

executions Availability: it refers to th e rate o f Serv ice

Activity ( SA). Av ailability ( A), during a t ime interval I, for a service s corresponds to:

A(s) = SA(s)/I (3) Scalability: it co mputes th e cap acity o f th e

service to manage loads. To test the scalability of a W S, we conduct ed the simulation while changing the number of concurrent clients.

Table 1: QoS Aggregation Functions per control statement

4. Our Simulation Framework SBAS

We have conduct ed si mulation experi ments usi ng NS-2 simulator [36]. NS-2 is a d iscrete-events simulator; its code is written in C++ with an OTcl in terpreter as a fro nt en d. NS-2 i s t argeted at net working research. It provides substantial support for si mulation of TC P, rout ing, and multicast prot ocols over wi red and wi reless networks. The main advantages of such an object-oriented sim ulator are reusability and easy m aintenance. To support SBA simulation, we have ext ended t he C ++ cl ass hi erarchy of NS-2 i n order t o i mplement HTTP, SMTP, and SOAP protocols. Our simulation framework (SBAS) is modular and includes: a graphical user interface, a BPEL generator, a simulation model generator, a m odels library and NS-2 si mulator. The

architecture of SBAS is presented in figure 6. User specifies the SB A under st udy. SB AS const ructs correspondi ng BPEL m odel. Sim ulation m odel is im plemented as actio ns organized in the object hierar chy o f th e NS-2 simulator. When simulation has been completed, results are collected and subjected to output QoS analysis.

5. Experimentations

In th is sectio n, we d escribe th ree SBAs wh ich satisfy the same funct ional requi rement. We use our di screte-event simulation approach to evaluate QoS of each of these SBAs.

Fig. 6 Our Framework for Simulating Service-Based Applications: SBAS

To express variability m odeling of SBAs, we adopt the MAP formalism [37]. A map i s a l abeled di rected graph with i ntentions as nodes and st rategies as edges between intentions. An intention is a goal that can be achieved by the performance of t he process. Each m ap has two distinct intentions Start and Stop to respect ively begin and end t he navigation into the map. A strategy is an approach, a manner to achieve an in tention. Th e MAP p ermits to cap ture variability by focusing on th e strateg y to ach ieve an intention and the potential a lternatives to accom plish the same intention. We consider an abstract intention Buy books online. Maps (described in Fi gure 7, Fi gure 8, and Fi gure 9) present possible refinements of t his abst ract intention. They model different SBAs (SBA1, SBA2, and SBA3) that ensure the same functional requirement Buy books online.

Aggregation function

Response Time (RT)

Reliability (R)

Availability (A)

Scalability (S)

Sequence n

i isRT1

)( n

i isR1

)( n

i isA1

)( n

i isS1

)( Switch

n

i ii sRTp1

)(* n

i ii sRp1

)(*

n

i ii sAp1

)(* n

i ii sSp1

)(*

Flow

niisRTMax ..1)(

n

i isR1

)( n

i isA1

)( n

i isS1

)(

Loop

)(* sRTk ksR )( ksA )( ksS )(


8

Fig. 7 Map of SBA1

SBA1 begins by invoking SearchByISBN (SI) service. This service allows the customer to search a book by entering its ISBN code. The to tal p rice to p ay is calcu lated u sing th e CalculateTotalPrice (CTP) service. The cl ient’s account i s then checked for suffi cient funds using the CheckCredit (CCr) service. If the client has sufficient credit, the ReleaseOrder (RO) service is invoked in order t o send t he book. Ot herwise, t he SendCreditLowInfo (SCLI) service is invoked.

Fig. 8 Map of SBA2

In SBA2, the client begins also by searching the book t hat he wants to buy using the SearchByISBN (SI) service. The FindLowestFare (FLF) service allows him to find the cheapest bookst ore. The cl ient’s credi t i s t hen checked by VerifyCredit (VC) service. If the client has sufficient credit, ChargeCard (CC), DispatchBook (DB ), and SendConfirmationSMS (SC S) are i nvoked. Ot herwise, InsufficientCreditSMS (ICS) is ex ecuted in order to inform the client of his insufficient credit.

Fig. 9 Map of SBA3

SBA3 begins al so by i nvoking t he SearchByISBN (SI) service. The FindPayPalBookStores (FPBS) service allows the client to find book stores accepting PayPal payments. The cl ient’s Pay Pal account i s checked by CheckPayPalAccount (CPA) service. If the client has a PayPal account , t he ChargeCard (CC) an d the ReleaseOrder (RO) services are i nvoked. Ot herwise, t he CreatePayPalAccount (CrPA) is ex ecuted to p ermit to th e client to create a PayPal account. Figure 10 shows the execution context of the user. Devices: PC, CPU: 1.86GHz, RAM: 2Go. Internet connectivity: NUMERIS, Mo dem sp eed:

512KB/s. Network: topology: user is connected to an Internet

Service Pro vider (ISP), wh ich is in its tu rn connected to Server through a Router (R).

– Link User ISP: t hroughput: 512KB /s, delay: 50ms.

– Link ISP R : t hroughput: 1MB/s, delay: 25ms.

– Link R Server: t hroughput: 512KB/s, delay: 50ms.

Fig. 10 User’s Execution Context

Table 2 illustrates the user’s preferences. For each QoS property i, the user attribute a weight.

Table 2: User’s QoS Preferences

QoS property Weight Response Time 0.35

Reliability 0.3 Availability 0.2 Scalability 0.15

The configuration of the simulation platform is Dual-Core based W indows XP system . For each QoS property, we performed a set of si mulation experiments, and we have considered the average value. Simulation results (figure 11, figure12, fi gure 13, and fi gure 14) show t hat SB A1 i s more reliable, available, and scalable than SBA2 and SBA3, but SBA2 is faster than SBA1 and SBA3.


9

Fig. 8 Response Time Simulation Results

Fig. 9 Reliability Simulation Results

Fig. 10 Availability Simulation Results

Fig. 11 Scalability Simulation Results

To validate these results in order to help the user to choose the appropriate SBA, we use a B enefit Function (BF). BF is computed as follows:

1*1 1

'

n

i

n

i iii withdBF (4) Where d'i is a norm alized value of a QoS dimension (i.e., QoS propert y) di and wi denot es t he user’s assigned relative i mportance t o t he di mension. As di mensions can be of di fferent units (e.g., response t ime is in second and availability in percentage), in order to allow for a uniform measurement of W S QoS i ndependent of units, data normalization is ap plied, wh ich essen tially tran sforms values of different units i nto com parable ones. B y considering a 75% confi dence i nterval, t he di mensions that are stronger with larg er values (e.g., reliability, availability and scalability) are norm alized according to the following equation:

otherwised

dmdd

ddmdifd

ddmdifd

nf

ii

ii

ii

5.0)(*4)(

)(*2)(0)(*2)(1

)(

'

'

'

(5)

While for QoS di mensions that are st ronger wi th smaller values (e.g., response time), they are normalized according to the following equation so that smaller values contribute more to the user’s benefit:

otherwised

dmdd

ddmdifd

ddmdifd

nf

ii

ii

ii

)(*4)(5.0

)(*2)(1)(*2)(0

)(

'

'

'

(6)

Where di i s t he val ue of di mension d for the service instance i, and m(d) and δ(d) are the m ean and standard deviation values for dimension d respectively. The validation of SBA1, SBA2, and SB A3 has gi ven the results shown in table 3. This validation proves that SBA1 is more appropriate for users’ expect ations than the SBA2 and SBA3 are.

Table 3: Validation Results

SBA1 SBA2 SBA3 BF 0.31 0. 265 0. 12


10

5. Conclusion and Future Work

In t his work, we adopt ed a discrete-events simulation approach to evaluate QoS of SBAs. W e presented a simulation modeling approach. Thi s approach enabl es an analytical description of SBAs and allows QoS predictions in t he di fferent st atus and condi tions of the execution context. We defined a light-weight quality model considering a set of QoS propert ies t hat can be measured by simulation techniques. We proposed al so a cont ext m odel t hat describes execution envi ronment and t he user’s preferences. Th is m odel is d epicted into the simulation model in order t o provi de a cont ext-based approach for evaluating SBAs. To show t he effect iveness of our approach, we have conducted a set of si mulation experi ments i n order t o evaluate and to validate three SBAs that provide the same required functionality. One possible extension of our work i s t he support of dynamic adaptations of SB As. It requi res ext ensive simulation experiments to define, validate and enhance the adaptation strategies.

Acknowledgments

This work has been support ed by t he European Communitys Sevent h Fram ework Program me FP7/2007-2013 under grant agreem ent 215483 (S-C ube). (http://www.s-cube-network.eu/). References [1] M. N. Huhns, and M. P. Singh, "Service-oriented Computing:

Key Concepts and P rinciples", in IEEE Internet Computing, Vol. 9, No. 1, 2005, pp. 75–81.

[2] F. Curbera, M. Duftler, R. Khalaf, W. Nagy, N. Mukhi, and S. Weerawarana, "Unraveling the W eb S ervices W eb: An Introduction to SOAP, WSDL, and UDDI", IEEE Internet Computing, Vol. 6, No. 2, 2002, pp. 86–93.

[3] M. P. Papazoglou, "Service-oriented Computing: Concepts, Characteristics and Directions", in Proceedings of WISE '03: International Conference on Web Information Systems Engineering, 2003, pp. 3–12.

[4] J. O’Sullivan, D. Edm ond, and A. Hofstede, "What’s in a Service? Towards Accurate De scription of Non-functional Service properties", Distributed and Parallel Databases, Vol. 12, No. 2, 2002, pp. 117–133.

[5] D. A. Menascé, "QoS Issues in Web Services", IEEE Internet Computing, Vol. 6, No. 6, pp. 72–75.

[6] J. Cardoso, J. Miller, A. Sheth, and J. Arnold, "Modeling Quality of Service for W orkflows and W eb Service Processes", Journal of Web Sema ntics, Vol. 1, No. 3, 2004, pp. 281-308.

[7] M. Sampath, R. Sengupta, S. Lafortune, K. Sinnamohideen, and D. Teneketzis, "Diagnos ability of Discrete-Event

Systems", IEEE Transactions on Automatic Control, Vol. 40, No. 9, 1995, pp. 1555–1575.

[8] M. Cordier, and S. Thiéba ux, "Event-based Diagnosis for Evolutive Systems", in P roceedings of DX '94: International workshop on Principles of diagnosis, 1994, pp. 64-69.

[9] Universal Description, Discovery and Integration specification 3.0.2, http ://uddi.org/pubs/uddi-v3.0.2-20041019.htm

[10] Web Services Description Language 2.0, http://www.w3.org/TR/wsdl20/

[11] Simple Object Access Protocol 1.2, http://www.w3.org/TR/SOAP/

[12] L. J. Osterweil, "Strategic Directions in Software Quality ", ACM Computing Surveys, Vol. 28, No. 4, 1996, pp. 738-750.

[13] G. J. My ers, "Art of Soft ware Testing", John Wiley & Sons (Eds.), ISBN: 978-0-471-04328-7, 1979.

[14] L. Baresi, and E. DiNitto, "Test and Analy sis of W eb Services", Elisabetta (Eds.), ISBN: 978-3-540-72911-2, 2007.

[15] S. Nakajima, "Model Check ing Verification for Reliable Web Service", in Proceedings of OOPSLA '02: Workshop on Object-Oriented Web Services, 2002, pp. 20.

[16] G. Salaün, L. Bordeaux, a nd M. Schaerf, " Describing and Reasoning on Web Services usi ng Process Alge bra ", in Proceedings of ICW S '04: IEEE International Conference on Web Services, 2004, pp. 43.

[17] H. Foster, S. Uchitel, J. Magee, and J. Kramer, "LTSA-WS: A Tool for Model-based Ve rification of Web Service Compositions and Choreography ", in P roceedings of ICSE '06: International Conference on Software Engineering, 2006, pp. 771– 774.

[18] L. J. Osterweil, "Form al Analy sis of W eb Service Compositions", Ph.D. Dissertation, 2007.

[19] R. Kazhamiakin, M. Pistore, and L. Santuari, "Analysis of Communication Models in Web Service Compositions", in Proceedings of WWW’06: International Conference on World Wide Web, 2006, pp. 267-276.

[20] A. Keller, and H. Ludwig, "The WSLA Framework: Specifying and Monitoring Service Level Agreements for Web Services", Journal of Network and Sy stems Management, Vol. 11, No. 1, 2003, pp. 57–81.

[21] H. Ludwig, A. Dan, and R. Kearney , "CREMONA: An Architecture and Library for Creation and Monitoring of WS-Agreements", in Proceedings of ICS OC '04: IEEE International Conference on Service Oriented Computing, 2004, pp. 65–74.

[22] F. Curbera, M. J. Duftler, R. Khalaf, W. Nagy , N. Mukhi, and S . W eerawarana, "Colombo: Lightweight Middleware for Service-Oriented Computing", IBM Systems Journal, Vol. 44, No. 4, 2005, pp. 799-820.

[23] L. Baresi, and S. Guinea, "Towards Dynamic Monitoring of WS-BPEL Processes", in P roceedings of ICS OC '05: International Conference of Service-Oriented Computing, 2005, pp. 269–282.

[24] L. Baresi, D. Bianculli, C. Ghezzi, S . Guinea, and P. Spoletini, "A Tim ed Extens ion of WSCoL", in Proceedings of ICWS '07: IEEE International Conference on W eb Services, 2007, pp. 663–670.

[25] S. Narayanan, and S. A. McIlraith, "Simulation, Verification, and Automated Composition of Web Services", in


11

Proceedings of WWW ' 02: International Conference on World Wide Web, 2002, pp. 77–88.

[26] S. Chandrasekaran, J. A. Mille r, G. A. Silver, I. B. Arpinar, and A. P . S heth, "Performance Analy sis and Simulation of Composite Web Services", Electronic Markets, Vol. 13, No. 2, 2003.

[27] E. Mancini, U. Villano, M. Rak, and R. Torella, "A Simulation-based Framework for Autonomic Web Services", in P roceedings of ICP ADS '05: IEEE International Conference on Parallel and Distributed Sy stems, 2005, pp. 433–437.

[28] M. Nikolaidou, and D. Angnostopoulos, "An Application-oriented Approach for Distributed System Modeling and Simulation", in P roceedings of ICDCS '01: International Conference on Distributed Computing Sy stems, 2001, pp. 165.

[29] M. Driss, Y. J amoussi, and H. Hajjami Ben Ghézala, "QoS Testing of S ervice-Based Applications ”, in Proceedings of IDT '08: IEEE International Design and Test W orkshop, 2008, pp. 45-50.

[30] S. Chen, B. Yan, J. Zic, R. Liu, and A. Ng, "Evaluation and Modeling of Web Services Performances", in Proceedings of ICWS '06: IEEE International Conference on W eb Services, 2006, pp. 437-444.

[31] Business Process Execution Language for Web Services 2.0, http://www.oasisopen.org/committees/tc_home.php?wg_abbrev=wsbpel

[32] D. Bertsekas, and R. Gallager, "Data Networks ", P rentice Hall (Eds.), ISBN: 978-0-132-00916-4, 1992.

[33] M. Keidl, and A. Kemper, "Towards Context-Aware Adaptable Web Services", in Proceedings of WWW ' 04: International Conference on World Wide Web, 2004, pp. 55–65.

[34] M. Driss, Y. Ja moussi, J. M. Jézéquel, H. Hajjami Ben Ghézala, "A Dis crete-Events S imulation Approach for Evaluation of S ervice-Based Applications ", in Proceedings of ECOW S '08 : IEEE European Conference on Web Services, 2008, pp.73-78.

[35] L. Zeng, B. Benatallah, A. Ngu, M. Dumas, J. Kalagnanam, and H. Chang, "Qos-Aware Middleware for W eb S ervices Composition", IEEE Transactions on S oftware Engineering, Vol. 30, No. 5, pp. 311–327.

[36] The ns manual, http://www.isi.edu/nsnam/ns/doc/index.html [37] C. Rolland, and N. Prak ash, "Bridging the Gap between

Organizational Needs and ERP Functionality", Requirements Engineering, Vol. 5, No. 3, 2000, pp. 180-193.

Y. Jamoussi received the Engineering degree from the Computer Engineering Faculty of the University of Tunisia, Tunis, in1989, and the Ph.D. degree from the same Faculty, in 1998. He is currently an Assistant Professor at the National School of Information Sciences, University of Manouba, Manouba, Tunisia. His current research interests focus on the enactment, guidance and the monitoring of strategic process. He is a co-author of a book on Method Engineering. He is an MVP on Biztalk. Recently, his research interests include web services composition.

M. Driss received an engineer degree in computer science in 2006 with distinction (Major of promotion) and master degree in software engineering in 2007 from the National School of Computer Science (ENSI), University of Manouba, Tunisia. She is currently a permanent researcher in the laboratory RIADI-GDL, ENSI, University of Manouba, Tunisia, and in the INRIA team-project Triskell, University of Rennes I, France. Her research interests include Web services composition, QoS of Web services, and QoS assurance techniques.

J-M.Jézéquel received an engineering degree in Telecommunications from the ENSTB in 1986, and a Ph.D. degree in Computer Science from the University of Rennes, France, in 1989. He first worked in Telecom industry (at Transpac) before joining the CNRS (Centre National de la Recherche Scientifique) in 1991. Since October 2000, he is a Professor at the University of Rennes, leading the INRIA research team Triskell. His interests include model driven software engineering based on object oriented technologies for telecommunications and distributed systems. H. Hajjami Ben Ghézala received Ph.D. degree from the Computer Engineering Faculty of the University of Tunisia, Tunis, in 1987. She is a Professor of Software Engineering at the National School of Information Sciences, University of Manouba, Manouba, Tunisia, leading the RIADI laboratory. Since the beginning of 2009, she is the rector of the University of Manouba. Her interests include Service Oriented Architecture.


12

Sentence Recognition Using Hopfield Neural Network

Bipul Pandey1, Sushil Ranjan2 , Anupam Shukla3 ,and Ritu Tiwari4

1Soft Computing and Expert System Laboratory, Indian Institute of Information Technology and Management Gwalior, Gwalior, Madhya Pradesh ,India




Abstract Communication in natural langua ges between computational systems and hum ans is an area that has attracted res earchers for long. This type of communication can have wide ramification as such a s ystem could find wide usage in several areas. Web-Browsing via input given as textual commands/sentences in natural languages is one such area. However, the enormous amount of input that could be given in natural languages present a huge challenge for machine learning of sy stems which are required to recognize sentences having similar meaning but different lexico-grammatical stru ctures. In this paper, we describe how a binary recurring neural network can be used to sufficiently solve this problem for Englis h. The s ystem uses the Hopfield Neural Network to rec ognize the meaning of text using training files with lim ited dictionary . Detailed analy sis and evaluation show that the system correctly recognizes/classifies approximately 92.2% of the input sentences according to their meaning. Keywords: Artificial Neural Network, Expert Systems, Machine learning, Natural Language, Sentence Recognition.

1. Introduction

Communicating wi th computer-based systems via natural language has l ong been consi dered t he fut ure of hum an-computer sy stem i nteraction. Though t he variety of such systems and their different requirements has slowed down the progress i n this field, the foremost problem is how to perform knowl edge Discovery from nat ural l anguage based data sources, whi ch specifically proves t o be m uch more difficult and prone t o errors. The ab ility o f Natural Languages t o create expressions with different arrangements of same or different words and phrases gives

rise t o l arge num ber of lexico-grammatically diverse sentences which may yet m ean sim ilar. Hence system discovering knowledge from such sources requi re such machine learning that takes care of the complex and immensely i mportant t ask of nat ural l anguage text processing in an opt imally su fficient way. The branch of Machine Learning i s concerned wi th desi gn and development of such procedures t hat make it possible for systems to cope wi th di verse dat a pat terns and change themselves accordingly. It is focused to m ake the systems to automatically adapt t hemselves t o recogni ze com plex patterns and make decisions accordingly. The basic problem to be sol ved is to successfully classify and recogni ze com mands i n t he form of sent ences in natural l anguages (Engl ish i n our sy stem). As such, our problem could also be seen in view of the much wider and immensely researched fi eld of Pat tern R ecognition and Classification. A few probl ems where t his approach has been im mensely successful are recognition and categorization of sounds, images, texts, features etc. [1]. A Pattern can be described as “opposi te of a chaos; i t i s an en tity, v aguely d efined, th at co uld b e given a name” [2]. A classification system is one that actually maps input vectors to a specific class. Hence, classification is basically t he job of l earning t he procedure that maps the input data [3]. A Pattern Recognition system basically classifies t he i nput pat terns. Thi s m ay be supervised or unsupervised dependi ng on t he input dat a and t echnique involved. The classes are defi ned usi ng pri or knowl edge about t he dat a rel ated dom ain [4] . The proposed system


13

uses Hopfield Classifiers to classify sentences according to their meaning. A Hopfield Classifier is a special type of Artificial Neu ral Netwo rk wh ich wo rks recu rsively to settle its output on one of the design points. These Design points are generally those on which the network settles for the initial training vectors. The network undergoes an unsupervised learning procedure. The System comprises of the data source, C oding module to create patterns of words and sentences according to the specified algorithm and Hopfield Networks for classifying words and sentences. The C oding M odule t akes i nput from t he dat a source for t raining bot h t he W ord Recognition Module and th e Sentence Recognition Module. The t raining i nput are i n t he form of sentences which are broken i nto separat e words. The separated words are coded into m atrix of size (N x M) each, where ‘N’ is the number of alphabets in the word and ‘M’ is total number of English al phabets. These coded pat terns of words are fed as t raining dat a t o t he W ord R ecognition Module. For Sentence Recogn ition Module, the training data comprises of sent ences t aken from t he dat a source whose i ndividual words are separat ed and coded by t he Coding Module in the same way . Then, t he m atrixes for each word per sentence are appended to form the sentence patterns in th e fo rm o f a co lumn m atrix an d fed as th e training patterns for t he W ord R ecognition M odule. The remaining sentences in the data source are then taken as a testing set. These are separat ed and coded i nto words similarly by t he codi ng m odule. The i ndividual word patterns are classified by the Hopfield classifier in th e Word R ecognition M odule t o si milar word training patterns. The word t raining patterns are t hen appended as per t heir posi tion i n t he sent ence t o gi ve t he sent ence pattern which is classified by the Hopfield classifier in the Sentence Recognition Module. Several genres of t echniques have been successful ly used for t he probl em of Pattern classification. However, Artificial Neural networks have st arted to gain a foot hold and are increasingly attracting researchers to use them to find novel y et effi cient m ethods of C lassification. The most popularly used ones are t he Feed-Forward Networks like Multi-Layer Percep trons. An other wid ely u sed network is Self-Organizing Map (SOM). Researchers have been enthusiastically applying the existing t echniques of Pat tern R ecognition and t he Artificial Neural Network paradigm to cater the problems of character, and word sentence recogni tion i n t extual, image and sound dat a. In [5] , an al gorithm had been proposed for the recognition of isolated off-line words. The al gorithm i s based on segm ent st ring matching and could do wi th moderately noisy and error prone dat a set .

In t heir paper, Jones et al [6] have talked about how Language Modeling can be used t o solve the problem of sentence reco gnition. Th ey h ave used probabilistic grammar along wi th a Hi dden M arkov Ident ifier for adequately co mpleting th is task . In [7 ], an algorithm had been proposed t o descri be a framework for classifier combination in gram mar-guided sent ence recogni tion. Hybrid Techniques have al so been used for the aforesaid problem. In [8] , Hi dden M arkov M odel (HM M) and Neural Network (NN) M odel have been combined for the solution. Here, W ord Recognition had been using a Tree-Structured dictionary while Sentence Recognition is done using a word-predecessor condi tioned beam search algorithm to segment into words and word recognition. In [9], sentence recognition has been achi eved which uses a template based pattern recognition and represents words as a series of di phone-like segm ents. In [10] , word co-occurrence p robability h as b een u sed fo r sen tence recognition. The incurred results were also compared with the m ethod usi ng t he C ontext Free Grammar. Binary Neural Networks have al so been successfully used in the task of pat tern recogni tion. B inary Ham ming Neural Network has been applied to recognize sentences and have been found to sufficiently succe ssful in this regard. The system proposed al so takes advantage of great er speed of the Binary Networks to provide a very efficient solution to the probl em of sent ence recogni tion [11]. David and Rajsekaran have talked about how Hopfield classifiers can be used as a tool i n Pat tern C lassification [12] . A combination of Hopfi eld Neural Network and Back Propagation Approach has al so been used t o propose method of vehicle license character recognition supported by a st udy of t he rel ations am ong t he st udy rate, error precision and nodes of the hidden layer [13].

2. Method Adopted

The architecture of the proposed system is shown in Fig 1. The first module comprises of t he selected database. This data base contains sentences in English extracted from any general text, for example, for an intelligent web-browsing system, it comprises of sentences/ com mands used while browsing the W orld W ide W eb (www) in text format. Hence, the database is dom ain specific, i.e. relevant to the sentences that will be classified using the system. First the sentences or com mands are extracted from the text database. The architecture of the proposed system is shown in Fig 1. The first module comprises of t he selected database. This data base contains sentences in English extracted from any general text, for example, for an intelligent web-browsing system, it comprises of sentences/ com mands used while browsing the W orld W ide W eb (www) in text format.


14

Hence, the database is dom ain specific, i.e. relevant to the sentences that will be classified using the system. First the sentences or com mands are extracted from the text database.

Fig. 1. Architecture of the System

Fig. 2. Format for Binary Coding of Isolated Words

The extracted sentences are then passed on t o t he W ord Isolation and Coding M odule. Thi s m odule separat es t he different words present in each sentence and codes them in the form of a m atrix wh ose fo rmat is sh own in th e Fig 2.The words are coded in a binary manner (using 1 & -1). This is done because binary inputs work m ore efficiently in the recurring neural net work bei ng used. Al so, use of non-binary inputs may lead to undesirable, spurious stable points i n t he net work. The words are coded keeping in mind the relativ e p osition o f th e alp habets in th e wo rd, denoted by t he row num ber i n whi ch t hey are present . What alphabet is present is represented by the colum n number in which they are presen t. Each column represents an al phabet of Engl ish dependi ng such that the alphabet represented by a column has t he sam e posi tion i n t he English alphabets as the col umn num ber of t hat col umn like th e 3 rd col umn woul d represent ‘c’ al phabet of English, where ‘c’ is at 3 rd al phabet i n Engl ish al phabet. For Example, the word “web” is coded as shown in the Fig 2. The fi rst ‘w’ i s represent ed by t he binary input 1 present in (1, 23) position in the matrix. Rest all positions are filled with the binary input -1 for the first row. Hence, for nth row all positions are -1 except where an alphabet is present Considered one of t he hi storical Neural Networks, the Hopfield Neural Net works are a t ype of Bidirectional Associative Memory. The network used here is the discrete versi on, i .e. t he uni t out puts are a di screte functions of the inputs.

Fig. 3. Architecture for Hopfield Neural Network


15

Being recurrent m eans the output of each node is drawn back to all other nodes via weights. The Hopfield network is created by suppl ying i nput dat a vect ors, or pat tern vectors, corresponding to t he di fferent cl asses. These patterns are called exem plar class patterns. In an n-dimensional data space the class patterns should have n binary com ponents {1,-1}; th at is, each class pattern corresponds t o a corner of a cube in an n-dimensional space. The network is then used to classify distorted patterns in to th ese classes. W hen a distorted pattern is presented to the network, then it is associated with another pattern. If t he net work works properl y, t his associ ated pattern is o ne of the class p atterns. The matrix of weight ‘W’ of a discrete time Hopfield classifier is defined:

)(/11

i

D

i

T

inW

(1)

where D is the number of exemplar class patterns { , ..., } vectors consisting of +/-1 elements, to be stored in the net work, and n i s t he num ber of com ponents, the dimension, of the class pattern vectors. Discrete-time Hopfield networks h ave th e fo llowing dynamics: ))(.()1( txWsigntx (2)

which is applied to one state x(t) at a tim e. At each iteration, the state to be updat ed is chosen randomly. The update continues till the netw ork converges. A distorted pattern, x (0), is used as initial state which is updated till convergence of the network to give the output which is the associated pattern.

Fig. 4. Format for Binary Coding for Sentences using recognized word pattern in column matrix form.

After this, t he coded words are passed on t o t he word recognition m odule. Thi s m odule com prises of a single layer recurri ng Hopfi eld Neural Net work trained with certain chosen words from th e d atabase. Th e Ho pfield Classifier present cl assifies t he i nput word patterns and gives as its output the training word pattern most similar to that input word. The di fferent words recogni zed for t he input sentence pattern are now com bined to give a pat tern in the sentence coding module which is the input for the sentence recogni tion m odule (Fi g 4). Thi s is done by appending the output patterns according to the position of the words in the sentence to give t he sent ence pat tern i n the column matrix form. This module also comprises of a single l ayered recurrent Hopfi eld Neural net work trained with chosen sent ence pat terns. The network iterates and converges t o gi ve out put whi ch is the training sentence pattern most similar to input sentence pattern. This output is th en m atched with th e o riginal sentence to check whether the output conveys the same meaning or not.

3. Experimental Results

The Database, used for constructing the proposed sy stem, is com prised of 500 sent ences, wi th 50 sent ences t o be used as t raining dat a for t he net work i n t he sent ence recognition m odule. The words of the 50 training sentences al ong wi th 100 m ore words were used for training t he net work i n word recogni tion m odule. The Database here was prepared by taking 500 sentences from a general text regardi ng com mon web usage. These sentences were t aken as a paragraph i n the form of a t ext file. The different sentences separat ed by ful l st ops were segregated using the Word Isolation and Coding Module. This module is built using Java Coding. Each sentence had its end marked by a full stop. The code takes advantage of this and separates sentences which start after o ne full stop and end before t he next one. The m odule t hen i solates each word per sentence and form s word pattern for each using the m atrix format described and outputs each in the form of a text file.

Table I. Details of Hopfield Classifiers Used Model No. of Layers

(word recognition

module, sentence

recognition module)

No. of Neurons

(word recognition

module, sentence

recognition module)

Transfer Function

Hopfield Classifier

(single, single) (393,500) Satlins

The 50 sentences chosen as the training data for the Hopfield Classifier in t he Sent ence R ecognition M odule


16

have their words along with 100 m ore words t aken as t he training set for the Word Recognition Module. The details of the Hopfield Classifiers us ed in each of the m odule are shown in Table I. The Sentence pattern of these 50 sentences is formed by appending the word patterns of each word of a sentence as per t heir posi tion i n t he sent ence t o gi ve t he sent ence pattern of that particular sentence. These patterns are then used as the training set for the Hopfield Classifier of the Sentence R ecognition M odule. The rem aining sent ences of the database are chosen as the testing data for the system. The words of each of the testing sentence are fed to the Word Recognition Module which gives the training word pattern most similar to that word as its output. The output patterns of each word are then appended as per its position to give the sentence pattern in the column matrix form as described earlier. The patterns of each of the sentence are then given to the Sentence Recognition Module where the Hopfield Classifier classifies and gi ves the most similar training sent ence pat tern as i ts out put. The output of each testing sent ence is then checked with the original testing sentence to find out where the output is similar in meaning it or not. Using t he proposed expert sy stem, we could recognize/classify approxi mately 92.2% of t he input sentences according to their meaning.

Table II. Result Obtained Technique

Used Total No. of Sentence Pattern In Database

No. of Training Sentence Patterns

No. of Training

Word Pattern

Performance

Hopfield 500 50 393 415/450

4. Conclusion and Future Works

The results show t hat t he proposed sy stem usi ng bi nary recurrent Hopfield classifier has been prom isingly successful to undertake sentence recognition. Though, these sentences are in differe nt l exical and gram matical format, yet the system has been able to successfully process the text in order to produce the desired results. The t ype of m odel descri bed i n the paper could be extended to developing other such exciting expert systems like an Intelligent Med ical System wh ich co uld tak e tex t input from patients and gives adequate prescriptions. One of t he m ain reasons for erroneous out put i n the proposed system turned out t o be t he m ismatch i n t he effective part of a word pat tern and that of the training words used. The effective part is the part where the binary

values turn out to be 1. Thi s means that that output of the word recognition is affected by the difference in the number of alphabets present in the input word and those in the training word pat terns. This sometimes grows over t o give i ncorrect sent ence m apping. Hence, this is an area where the proposed system could be worked upon. References [1] Paulus, D., and Hornegger, J., Applied Pattern Recognition,

2nd edition, Morgan Kaufmann Publishers, 1998. [2] Watanabe, S., Pattern Recognition: Human and Mechanical,

Wiley, 1985. [3] Weiss, S. I., and Kulikowski, C., Com puter S ystems That

Learn: Classification and Prediction Methods from Statistics, Neural Networks , M achine Learning and Expert Sy stems, Morgan Kaufmann Publishers, 1991.

[4] Han, J. and Kamber M., Data Mining : Concepts and Techniques, 2nd edition, Morgan Kaufm ann Publishers, 2008.

[5] Favata, J. T., “General Word Recognition: Using Approximate Segment-String Matching”, in 4th International Conference on Document Analy sis and Recognition, 1997, pp. 92-96.

[6] Jones, G. J . F., Wright, J . H.,Wrigley, E. N ., and Carey, M. J., “Isolated Word Sentence Recognition Using Probabilistic Context- Free Grammar”, IEEE Colloquium on Systems and Applications of Man Machine Interaction Using Speech I/O, 1991, pp. 13/1-13/5.

[7] Jiang, X., Yu, K., and Bunke, H., “Classi f er combination for grammar-guided sentence recognition”, in Multiple Classifier Systems, First In ternational Workshop, 2000, pp. 383–392.

[8] Marukatat, S., Artikres, T., Gallinari, P., and Dorizzi, B., “Sentence Recognition through Hybrid Neuro-Markovian Modeling”, in Sixth International Conference on Document Analysis and Recognition, 2001.

[9] Rosenberg, A. E., “Connected Sentence Recognition Using Diphone-like Templates”, in International Conference on Acoustics, Speech, and Signal Processing, 1988, vol.1, pp. 473-476.

[10] Murase, I., and Nakagawa, S., “Sentence Recognition Method Using Word Co-occurrence Probability and Its Evaluation”, in International Conference on Spoken Language Processing, 1990, pp. 1217–1220.

[11] Majewski, M., and Zurada, J. M., “ Sentence Recognition Using Artificial Neural Network”, Knowledge-Based Systems, vol. 27, issue 1 , 2008.

[12] David, V. K., and Rajase karan, S., Pattern Recognition Using Neural and Functional Networks, Springer, 2008.

[13] Qiu, M., Mai H., and Liao X., “Application of neural network in vehicle licens e character recognition”, Jisuanji Gongcheng yu Sheji (Computer Engineering and Design). Vol. 29, no. 8, 2008, pp. 2041-2043.


17

B. Pandey is a student of final year of 5-year Integrated Post Graduate Course (BTech + MTech in IT) in Indian Institute of Information Technology and Management, Gwalior, India. His areas of research are soft computing, artificial intelligence, hybrid system design and Biometrics. S. Ranjan is a student of final year of 5-year Integrated Post Graduate Course (BTech + MTech in IT) in Indian Institute of Information Technology and Management, Gwalior, India. field of research includes soft computing and Bioinformatics. A. Shukla is an Associate Professor in the ICT Department of the Indian Institute of Information Technology and Management Gwalior. He completed his PhD degree from NIT-Raipur, India in 2002. He did his post graduation from Jadavpur University, India. He has 22 years of teaching experience. His research interest includes Speech processing, Artificial Intelligence, Soft Computing and Bioinformatics. He has published around 120 papers in various national and international journals/conferences. He is referee for 4 international journals and he is in the Editorial board of International Journal of AI and Soft Computing. He received Young Scientist Award from Madhya Pradesh Government and Gold Medal from Jadavpur University. R. Tiwari is an Assistant Professor in the ICT Department of Indian Institute of Information Technology and Management, Gwalior. Her field of research includes Biometrics, Artificial Neural Networks, Speech Signal Processing, Robotics and Soft Computing. She has published around 60 papers in various national and international journals/conferences. She has received Young Scientist Award from Chhattisgarh Council of Science & Technology and also received Gold Medal in her post graduation.


18

Review Strategies and Analysis of Mobile Ad Hoc Network- Internet Integration Solutions

Rakesh Kumar1, Anil K. Sarje2 and Manoj Misra3

1 Research Scholar, Department of Electronics and Computer Engineering, Indian Institute of Technology, Roorkee, India–247667

2 Professor, Department of Electronics and Computer Engineering, Indian Institute of Technology, Roorkee, India –247667

3 Professor, Department of Electronics and Computer Engineering, Indian Institute of Technology, Roorkee, India –247667

Abstract

The desire to be connected anytime and anywhere has led to the development of wi reless net works, openi ng new vista of research i n pervasi ve and ubi quitous com puting. Mobile Ad Hoc Net works (MANETs) use portable devices such as mobile phones, laptops or personal digital assistants (PDAs) for spont aneous est ablishment of communication. M ost exi sting research in the area of mobile Ad Hoc Networks is limited to stand-alone isolated networks. But connectivity of a mobile Ad Hoc network to the Internet is also desirable as m ore and m ore applications and services i n our soci ety now depend on fixed infrastructure networks. It is therefore important that dynamically deployed wi reless Ad Hoc net works shoul d also gain access to these fixed networks and their services. The integration of MANETs in to In ternet in creases th e networking flex ibility an d co verage of existing infrastructure net works. Although researchers have proposed many solutions, but it is still unclear which one offer the best perform ance compared to the others. W hen an Ad Hoc network is connected to Internet, it is important for the mobile nodes to detect efficiently available Internet gateways providing access to the Internet. Internet gateway di scovery t ime and handover del ay have st rong influence on packet del ay and t hroughput. The key challenge i n provi ding connect ivity i s t o m inimize t he overhead of m obile IP and Ad Hoc rout ing protocol between Internet and Ad Hoc networks. There, this paper focuses on proposed t echnical sol utions on Int ernet gateway di scovery and al so we bri efly descri be di fferent ways to provide global Internet access for MANETs.

Finally, some challenges are also mentioned which need in depth investigation. Keywords: MANET, Internet Gateway Discovery, Mobile IP, Address Autoconfiguration, DAD, Internet, AODV

1. Introduction

A Mobile Ad Hoc Network [1] is an autonomous network that can be form ed wi thout t he need of any established infrastructure or centralized adm inistration. Vari ous routing protocols have been proposed for M ANETs [39] . But one drawback of M ANETs i s that communication is limited to t he Ad Hoc dom ain onl y. M any appl ications however need a connection to an external network, like the In ternet. As illu strated in Fig . 1 , in o rder to p rovide Internet connectivity to the nodes i n an Ad Hoc network, routers or one or m ore nodes i n the Ad Hoc network can serve as Internet gateway s to an external network, where the external net work can be an i nfrastructured net work such as LAN, Internet or a cellular network, or even an infrastructure-less net work such as anot her Ad Hoc network. When connecting M ANETs wi th t he Int ernet, the ro uting in teroperability b ecomes a cru cial ch allenge. Ad Hoc nodes can not obtain routing information beyond the sco pe o f th e MANET. Th erefore, th e in teroperability between IP routing and Ad Hoc rout ing needs to be given attention. W hen an Ad Hoc network i s connect ed t o t he Internet, i t i s i mportant for t he m obile nodes to detect efficiently available Internet Gateways providing access to the Internet. Internet gateway discovery time and handover delay have st rong i nfluence on packet del ay and


19

throughput. The key challenge in providing connectivity is to minimize the overhead of m obile IP [14,37] and Ad Hoc rout ing protocol between i nfrastructure and Ad Hoc networks.

Internet

Gateway

Mobile node

AdhocNetwork

Fig. 1 Mobile Ad Hoc network connected with Internet

The challenge in integration of MANET-Internet is to inform m obile nodes about avai lable Internet gateways while making a m inimal consum ption of t he scarce network resources. So, an effi cient gateway discovery for Ad Hoc net works becom es one of t he key el ements to enable t he use of hy brid Ad Hoc net works in future mobile and wi reless net works. In order t o be abl e t o communicate with the Internet , each m obile node within the MANET must configure globally rout able IP address [2]. Thi s i ncludes acqui ring a temporary address once it enters the MANET, wh ich allo ws o nly lo cal communication, within the Ad Hoc network. Next it needs to discover and sel ect one Int ernet gat eway t o use i ts prefix and form a gl obally rout able address. Thi s paper also tackles t his fundam ental probl em and addresses t he interworking between Ad Hoc net works and t he Internet. Let us take a closer look at the interconnection of MANET and fixed net work. Unl ike t he ful ly hi erarchical addressing scheme used i n the Internet, MANETs have a completely flat addressing model. In fact , Ad Hoc rout ing protocol such as AODV does not employ the concept of IP subnet. They assum e t hat a node i n a M ANET m ay use any IP address provi ded that i t i s not duplicated. In fact, Ad Hoc routing protocols use host -based rout es rat her than network p refixes. Un like in trad itional IP n etworks, two neighboring nodes are not required to have addresses belonging to the sam e IP subnet t o be abl e t o di rectly communicate with each other. These differences with traditional IP networks create some interworking issues as indicated below [3].

Discovering Internet Gateways Address Auto-Configuration Reaching a Destination Duplicate Address Detection (DAD)

This paper presents a surv ey of recent advances in technical issues in connecti ng MANETs to th e In ternet. We anal yze sol utions proposed by ot her researchers and describe our up-t o-date cont ributions. The remainder of

the paper i s organized as fol lows: In section 2, we briefly describe Internet routing protocols. The MANET-Internet connectivity, including the basi c prot ocol st ack i s described i n sect ion 3. Address aut oconfiguration and Internet gateway di scovery approaches are bri efly described in section 4 and 5 re spectively. In section 6, we present analysis of different In ternet co nnectivity proposals with mobile Ad Hoc net works. Sect ion 7 summarizes and com pares diffe rent t echnical sol utions proposed for M ANET-Internet i nterconnectivity and has been gi ven i n t abular form (Tabl e 1). Finally, section 8 concludes the paper and defines topics for future research.

2 Internet Routing Protocols

Routing i n Int ernet i s IP address based. Each IP address consists of network id and host id portion. Routing decisions are taken by rout ers for packet s based on t he network i d port ions of t he destination IP addresses [10]. The IP addresses of nodes wi thin t he same network t hus share t he com mon net work i d whereas t he node address portion of t he IP address i dentifies a specific node in the network. The highest l evel of t he Int ernet hi erarchy consists of a num ber of Aut onomous Systems. Each Autonomous System is a di stinct routing domain. Routers communicate with each other within an Autonom ous System usi ng i ntra-domain rout ing prot ocols, which are also known as Int erior Gat eway Prot ocols. Gat eway routers are used t o i nterconnect di fferent Aut onomous Systems. Exterior Gateway Protocols are used to exchange routing i nformation bet ween Aut onomous Systems. Routing i nformation Prot ocol (R IP) [12] and Open Shortest Path First (OSPF) [13] m ay be used as Interior gateway protocols. OSPF is a m ember of the “link state” family an d co mmonly u sed n owadays. It u ses multi-metrics of a l ink that may consider bandwidth, hop count , and reliability. A router in OSPF is aware of all links between al l rout ers of an Autonomous System. In this, routers m aintain a m ap of t he whole network that is updated if a change in t he net work t opology i s det ected. Based on this knowledge, routers in OSPF calculate shortest/best path from source to destination. Border Gat eway Prot ocol (B GP) [15] provi des connectivity bet ween di fferent Aut onomous Sy stems and also provi des shari ng of rout ing information by one Autonomous Sy stem wi th ot her Aut onomous Sy stem. It exchanges routing tables to other Autonomous Systems on demand.

3. MANET-INTERNET Connectivity

Whenever a MANET node i s t o send packet s t o a fi xed network, it must transmit the p ackets to a g ateway. A


20

gateway acts as a bridge between a MANET and the Internet. Therefore, it has to implement both the MANET protocol stack and t he TC P/IP sui te. In t he phy sical and data link layer, a mobile Ad Hoc node runs protocols e.g. (IEEE 802.11 DCF) that have been designed for wireless channels. In the network layer, either an IP based Ad Hoc routing prot ocol e.g., Ad Hoc On Dem and Di stance Vector Routing (AODV) protocol [20] is used, or this layer i s divided into two sub l ayers, namely the usual IP layer over a non IP based Ad Hoc routing protocol that transports the IP packets in th e ad n etwork in an encapsulated manner. Gateway contains protocols of both the fixed Internet and the wi reless Ad Hoc net work. On the Int ernet si de, i t runs t he usual Int ernet prot ocols. On the Ad Hoc side, it sends and receives packets using an Ad Hoc ro uting alg orithm. Mobility management is performed using Mobile Internet Protocol [14]. The basi c protocol st acks for m obile nodes, Internet gateways and Internet hosts are depicted in Fig. 2.

GATEWAYAPPLICATION

UDP/TCP

IP/Mobile IP

Wired PHY

INTERNET HOST

Wired LINK (e.g.IEEE 802.3)

APPLICATION

UDP/TCP

AODV (andpossibly Mobile IP)

LLC802.11 MAC

802.11 PHY

MOBILE HOST

APPLICATION

LLC802.11 MAC802.11 PHY

Wired Link (e.g.IEEE 802.3)

AODV IP/Mobile IP

UDP/TCP

Wired PHY

Fig 2. Protocol stack for MANET-Internet connectivity

The In ternet g ateway p rovides an illu sion to th e outside world that the MANET is simply a normal IP subnet.

4. Address Autoconfiguration

Once a m obile node has chosen one gat eway, all packets sent over t his Gat eway to the Internet need a source address with the sam e prefix as the Gateway. Ad Hoc node needs an address aut o-configuration mechanism in order to confi gure a gl obal rout able and t opological correct address i n order t o avoi d ot her solutions like Network Address Transl ation (NAT). IPv6 defi nes two fundamental principles for auto-configuration: stateful and stateless confi guration. In st ateful aut o-configuration [3] , the IP address of a node is assigned by a central entity (e.g., a Dy namic Host C onfiguration Prot ocol [DHCP] server resid ing in th e g ateway). It automatically assigns addresses to requesting mobile nodes and m anages t he address space. However, centr alized approaches are not suitable for MANETs due to possible network partitions,

although it has been consi dered i n som e works [16] . Another opt ion i s t o use stateless auto-configuration (addresses generat ed by t he nodes themselves) [17,18,9]. In th is ap proach, th e g ateway can advertise within its control messages a network prefix from which the nodes can deri ve an IP address. B y integrating the auto-configuration i nformation i nto gat eway di scovery messages, the overall overhead i s reduced. However, t he network m ight still be periodically flooded with prefix information, whi ch consum es preci ous resources. It is advisable t o confi gure t he nodes with an IP address belonging to the subnet of i ts defaul t gat eway. Thi s guarantees that the access router does not need to perform network address translation (NAT) to get messages routed back from the Internet. Once a node has an IP address, it may check whether other nodes use t hat address. If t wo mobile nodes are using the same address, then the address shoul d be deal located and the node should t ry to get another one. Thi s procedure i s known as Dupl icate Address Det ection (DAD) [19] and can be perform ed by aski ng t he whol e MANET if this address is already in use. W hen a node receives one of those messages requesting an IP address whi ch i t owns, then it replies to th e o riginator in o rder to n otify th e duplication.

5. Internet Gateway Discovery Approaches

For access to global services, an Internet Gateway (IGW) in the access network can provide Internet connectivity for nodes i n t he M ANETs. Mobile nodes from Ad Hoc network can use this route to send/receive packets addressed t o or from Int ernet. As mentioned before, t he st andard Ad Hoc rout ing prot ocols do not provide th e fu nctionality o f d etecting In ternet g ateways, thus the protocols have t o be ext ended. The ext ensions to the standard Ad Hoc rout ing prot ocols are based upon special Ad Hoc rout ing messages. The gateway discovery can be realized in three different way s: reactively, proactively, or in a hybrid way.

Reactive Discovery Approach

The react ive gat eway di scovery onl y provi des the gateway information if a part icular Ad Hoc node request s the gateway to get access to th e Internet. In this case, mobile node broadcasts a Gateway Solicitation (GWSOL) message within the entire Ad Hoc network. The GWSOL is pi ggybacked on rout e request (R REQ) message of MANET rout ing prot ocols. Int ermediate nodes rebroadcast this special route request message (RREQ_I). If the gateway receives the GW SOL, it sends special route reply (RREP_I) m essage back t o t he m obile nodes offering its services and IP a ddress (or IP prefi x address),


21

so t hat t he request ing node can set up a route to the gateway.

Proactive Discovery Approach

In this approach, every Internet gateways periodically broadcast their services and IP prefi x address t hroughout the MANET. Any mobile node that wants to interact with the Int ernet nodes det ects t his packet and begins registration. When a mobile node connected to an Internet gateway receives an advertisem ent from another Internet gateway, it m ay d ecide to co nnect to th e new Internet gateway, i f i t provi des a bet ter servi ce. The proact ive discovery mechanism reduces the average delay compared to the reactive discovery but incurs higher communication overhead.

Hybrid Discovery Approach

This uses a mixture of bot h t he above approaches, which provides a t rade-off i.e., between the advantages of proactive and react ive approaches, providing good connectivity while keeping the overhead costs low. In this approach, t he peri odical Int ernet gateway advertisements are not flooded throughout the whole Ad Hoc network but only sent t o m obile nodes t hat are i n t he vi cinity of the Internet gateway, i.e., the time to live (TTL) of those advertisement p ackets is lim ited. Nodes t hat are further away have to solicit advertisements reactively.

Two ki nds of i nformation are needed by MANET node for routing packets via an Internet gateway. The first one is the address of t he Internet gateway. To rout e data packets and cont rol messages to the Internet gateway, Ad Hoc nodes must be aware of t he gat eway’s Ad Hoc routable address. The second one i s, a default route pointing to the In ternet g ateway. All p ackets th at are destined to the Internet are routed vi a t he defaul t route. For g aining th e g ateway in formation, i.e. th e gateway’s address and route to the gateway, either the advertisement based (proactive) o r th e so licitation b ased (reactiv e) approach may be utilized.

6. Analysis of Different Internet Connectivity Proposals for MANETs

When the Ad Hoc network is interconnected to an IP network, mobile nodes in the Ad Hoc network need global addresses t o com municate wi th t he Int ernet and node mobility should be properly dealt with [21,22]. Especially, when mobile nodes move t o anot her area, t heir subnet changes and a new IP address must be obt ained. Several solutions have been proposed t o deal with the integration of M ANETs t o t he Int ernet. M ost of the proposed solutions require the addition of gateways and differ in the design and functionality of th e g ateways, n umber o f

occurrences, and the routing protocols used within the Ad Hoc network. Since Internet gateways have two interfaces they are part of the Internet and t he Ad Hoc net work simultaneously. They understand the Internet protocol (IP) as well as a M ANET routing protocol (e.g. AODV). Mostly, the existing approaches consi der onl y fi xed gateways to connect MANET nodes to the wired Internet. We bri efly di scuss sol utions for bot h fi xed Internet gateways and mobile Internet gateways [5].

6.1 Fixed Internet Gateway Approaches

Bin et al. [23] proposed an adapt ive gat eway discovery schem e that can dynamically adjust the TTL value of Agent Advertisem ents (GW ADV m essages) according to the m obile nodes to Internet traffic and the related position of m obile nodes from Int ernet gat eway with which they registered. This protocol provides Internet access to MANET m obile nodes using m obile IP [14,37]. The prot ocol uses forei gn agent s t o t rack and forward packets t o and from m obile nodes. Forei gn agent periodically cal culates t he average hops conveyed by RREQ_I m essage or regi stration request sent by mobile nodes requesting Int ernet connect ivity. So t he broadcast radius of Agent Advert isements can be adjusted dynamically according to real time demand for the Internet access and the factual network conditions.

Ratanchandani et al. [24] proposed a hybrid gateway discovery approach to discover gat eways t hat l imits t he effects of broadcast overhead. AODV and two Mobile IP [14,37] forei gn agent s are used to interconnect MANET and the Internet. However, th e TTL of the foreign agent’s advertisements is limited to onl y a few hops. Thus, onl y mobile nodes t hat are cl ose t o one of t he forei gn agent s receive the agent advertisements. Nodes that are further away h ave to so licit ad vertisements reactively. Intermediate nodes are al lowed t o repl y on a sol icitation with agent advertisem ents and to eavesdrop and cache agent advertisement information that is sen t by unicast to the request ing m obile node. The performance of this approach depends on the Tim e-To-Live (TTL) value, which is set for a part icular scenari o and net work condition under considerations. In order to switch between foreign agents, the MIPMANET Cell Switching algorithm [21] is used.

Hamidian et al. [27] gave a sol ution, which provides Internet connectivity to Ad Hoc net works by m odifying the AODV routing protocol. An “I” flag is added as an extension to AODV RREQ and RREP to locate the fixed node. If after one network-wide search without receiving any corresponding route replies, the mobile node assumes that the destination is a fixed node, which is located in the Internet and thus del ivers the packets through a gat eway. Three methods of gateway discovery for a m obile node to


22

access the Internet are provide d: proactive, reactive and hybrid approach. Al l of t hem are based on the number of physical hops t o gat eway as t he m etric for t he gat eway selection.

Rosenschon et al . [7] proposed a proactive gateway discovery m ethod i n whi ch gat eway peri odically sends HELLO messages that contai n a special option called PROAGW option. This opt ion has al l i nformation about the gateway that is n eeded to set u p a ro ute to it. All Ad Hoc nodes that have received the PROAGW option can add the option to th eir o wn h ello m essages. If m ultiple gateways are available, then mobile node receives multiple answers. The Ad Hoc node has to decide which gateway it should use. For sel ection of best Int ernet gat eway, parameters that can be consi dered are hop count , congestion, overl oad, avai lable bandwi dth, delay etc. In both approaches, the M ANET i s fl ooded wi th rout ing messages.

Sun et al. [25] di scussed t he perform ance of t he integration o f th e Ad Ho c On-Demand Distance Vector (AODV) routing protocol a nd M obile IP [14,37] . It presents a m ethod for enabl ing nodes wi thin an Ad Hoc network to obtain Internet connectivity when one or more nodes i s wi thin di rect t ransmission range of a foreign agent or m ore specifically an Internet Gateway/Access Router. In t heir approach, an Ad Hoc net work is connected to a foreign agent, which basically has the same functionality as an In ternet g ateway (IGW). Internet Gateway assigns a gl obal prefix for t he Ad Hoc network, which makes i t possi ble for m obile nodes i n Ad Hoc network to com municate with Internet. W hile AODV is used for route discovery and maintenance within MANET, Mobile IP [14,37] provi des m obile nodes with care-of-addresses. However, handoff occurs only if a mobile node has not heard from its forei gn agent for m ore t han one beacon interval, which is the time between two successive agent ad vertisements, o r its route to a foreign agent has become invalid.

Broch et al. [26] proposed a sol ution for t he integration of M ANET wi th M obile IP [14,37] using a source rout ing protocol. They introduced a border router, which has two interfaces. Rou ting on Internet gateway’s interface internal to the Ad Hoc network is accom plished using dynamic source routing (DSR) [35] protocol, while its interface connected to the In ternet is configured to use normal IP rout ing m echanisms. M obile nodes in an Ad Hoc net work are assi gned hom e addresses from a single network. The nodes wi thin range of t he foreign agent act as gateways between the Ad Hoc network and the Internet. As a reactive approach, foreign agent discovery is only done when required. Traditional IP routing is used on the Internet side while within MANET DSR p rotocol is u sed. Foreign agents are responsible for connect ing the Ad Hoc network with the Internet.

In [21] , Jonsson et al . proposed a method, called MIPMANET based on AODV [16] , but it provides Internet access by using tunneling and Mobile IP [14,37] with forei gn agent care-of addresses. Fi g. 3 depicts the layered architecture of Mobile IP [14,37] and Ad Hoc routing functionality. Mobile nodes that want Internet access register with a foreign agent and tunnel all packets destined fo r th e In ternet to the registered foreign agent. The packet s dest ined for t he Int ernet are t unneled to the foreign agents, which in turn fo rward th e p ackets to th e destination in the Internet. The host s that do not requi re Internet access see the Ad Hoc network as a standalone network. The t unneling approach al so enabl es MIPMANET to incorporate the default route concept into on-demand rout ing. The Ad Hoc on demand distance vector routing protocol AODV [20] is used within the mobile Ad Hoc net work and del ivers packet s bet ween mobile nodes and forei gn agent s. M IPMANET al lows a visiting node t o switch from its current foreign agent to a new one, a phenom enon known as handoff, onl y if it is at least two hops closer to the new one. It utilizes a new algorithm, called MIPMANET Cell Switch ing (MMCS), to det ermine when m obile nodes i n t he Ad Hoc network should register with a new foreign agent. In t his solution, it is assumed that a mobile node that wants Internet access has been assi gned a hom e address t hat i s val id on t he Internet [22]. Authors identified the benefi ts of usi ng the closest gat eway and proposed a gat eway sel ection algorithm based on hop count . A si mulation st udy indicates the benefits of broadcasting agent advertisements compared to using unicast solicitation/advertisement.

Transport

Mobile IP

IP

Correspondent NodesVisiting Nodes

Foreign AgentHomeAgent

Ad Hoc Network IP Network

Fig 3. MIPMANET conceptual view


23

Lee et al . [30] proposed a hy brid gateway discovery scheme and com pared it to a reactive one. It requires a source routing protocol in the Ad Hoc network. A gateway only sends out new advert isements when i t det ects any topology change i n t he Ad Hoc network. Moreover, advertisements are only forwarded to nodes that are either connected t o t he Int ernet or t hat have act ually moved. Advertisements are only generated if the ratio between the number of Int ernet joi ning nodes and t he num ber of advertisement forwarding node s exceeds a certain value. In addi tion t o t he adapt ive advertisements, conventional advertisements are broadcasted with a relatively long time interval. They rely on a source based rout ing prot ocol, which limits the applicability to particular type of routing protocol.

Tseng et al . [31] proposed a dy namic m obile agent service coverage schem e by i ntegrating DSDV [4] and Mobile IP [14,37]. They present ed a sol ution where MANETs are treated as Int ernet subnet s (Fi g. 4). The existence of each hom e agent (HA)/foreign Agent (FA) is known only up to N wireless hops away from the agents as the agent advert isement can onl y t raverse up t o N hops. Mobile nodes whi ch are m ore t han N hops away from HA/FA are req uired to b roadcast ag ent so licitation messages to search for a gateway. Similarly, the number of physical hop is used as the only metric to determine which ideal HA/ FA t he m obile node should connect with. The paper does not include any evaluation of t he system or a comparison to ot her sol utions. The proact ive approach gives a high overhead when mobility is high and the hop count gateway selection could cause problems.

I n t e r n e t

MANET1 MANET2

MANET4MANET3

G1 G2

G3 G4N=3

N=2

N=3

N=3

A

D

B

C

Fig 4. The proposed network architecture extending each access point to a MANET

Xie [40] et al . proposed an enhanced DSDV (EDSDV) protocol for mobile Ad Hoc networks for providing global bi-directional MANET-Internet connectivity. The EDSDV protocol integrates with Mobile IP [14,37] for support ing

bi-directional global connect ivity for m obile nodes by using Forei gn Agent as t he m obile IP proxy . Mobile IP protocol provides the global mobility for a m obile host to access Internet resources while visiting a foreign network.

FA

CN

HA

Internet

AccessPoint

12

54

6

7

8

3

Wireless Network

Mobile Node

HA : Home Agent

CN : Correspodent Agent

FA : Foreign Agent

9

Fig 5. Proposed MANET- Internet integration architecture

Fig. 5 shows how Mobile IP and Ad Hoc routing protocols coordinate for building connectivity across het erogeneous networks. The proposed prot ocol includes three components nam ely enhanced DSDV rout ing prot ocol, mobile IP proxy and connection management. Performance metrics chosen for proposed prot ocol evaluation are packet deliv ery, overhead and packet latency.

Wakikawa et al . [28] defi ne both proactive and reactive schemes, which are not dependent on any routing solution. They proposed an approach t o gl obal Int ernet connection over the IPv6 M ANET envi ronment, where mobile nodes in the Ad Hoc net work are confi gured with new globally routable IP addresses based on the neighbor discovery prot ocol (NDP) of IPv6 or route searching procedure of on-dem and rout ing protocol. This paper defines t wo di fferent m echanisms t o di scover Internet gateways: peri odic fl ooding of gat eway advertisement (GWADV) m essages from the gateways and reactive flooding a gat eway solicitation (GWSOL) messages from nodes. The gat eway advert isement m essage cont ains t he global IPv6 address of t he gat eway, t he net work prefi x advertised by th e g ateway, th e p refix len gth an d th e life time associated with th is in formation. They specify a stateless auto-configuration mechanism, which is based on network prefi xes advert ised by Int ernet gat eways. The nodes concatenate an interface identifier to one of those prefixes in order to generate the IP address.

In [29] , E.M . B elding-Royer et al . proposed Mobile IP, which was support ed by IPv4 Ad Hoc net works wi th AODV rout ing prot ocols. The proposed schem e has a proactive agent solicitation procedure with AODV route search t o regi ster t o M obile IP. It distinguishes the location of destination nodes using F-RREP of FA, when a packet is sen t to the In ternet. In addition, it is capable of packet routing using default routing of FA. However, t his


24

proposal does not consider the selection between m ultiple FAs. Also, it delays the connection setup time because this proposal first needs to conclude that the destination is not within the Ad Hoc net work before a m obile node can use the FA.

C. Jelger et al. [32] proposed a proact ive approach in which Internet gateways p eriodically ad vertise th eir presence by flooding information (GW_INFO) messages. This proposal uses a rest ricted flooding scheme, which is based on the idea of prefix continuity to limit the overhead of the proactive gateway discovery. The prefi x continuity guarantees that every node shares t he sam e prefi x and each gateway only receives IPv6 data packets belonging to its prefi x. A m obile node sel ects onl y one of the GW_INFO messages according to some metrics. Then the node configures an IPv6 address based on the advertised prefix and sends onl y t he GW _INFO i ncluding t he selected prefix. However, if th e approach is unified with a reactive routing protocol, then a node i n the network must discover a rout e ot herwise i t causes a break of the connection because of the feat ure of the reactive routing protocol. They speci fy a stateless auto-configuration mechanism, which is based on network prefixes advertised by gateways. The nodes concat enate interface identifier to one of those prefixes in order to generate the IP address. A mobile node sel ects t he best pat h t owards t he gat eway using one o f th e m etrics su ch as d istance, stab ility, o r delay from all the gateway information messages received.

In [3 4] th e scalab ility o f b oth ap proaches (p roactive and reactive) is com pared w ith resp ect to th e n umber o f Internet Gateway s by Ghassemian et al. The fixed access network to gether with th e Ad Ho c frin ge constitutes a multihop access network as depicted in Fig. 6.

IGW

IGW

IGW

MN

MultihopAccess

Networks

Internet

Fixed Access Network

Fig. 6 Multihop access network

The simulation results show that the proactive approach is more advantageous because the packet delivery ratio is higher and, although the signaling overhead i s larger too, it i s reduced for a hi gher num ber of Int ernet gat eways, because the am ount of periodical gateway advertisements is increased b ut m ore d ata p ackets are tran smitted successfully. The hybrid Int ernet gat eway di scovery approach is also com pared that shows the average packet delay and t he packet del ivery rat io. The hybrid gateway discovery represents a balance between the reactive and the proactive approaches.

In [8], AODV routing protocol for Ad Hoc networks has been modified to offer enhanced Internet connectivity and then in depth investigation has been carried out on the three Int ernet gat eway di scoveries i.e. reactive, proactive and hy brid for provi ding i nter-connectivity between Ad Hoc networks and Internet. The performance metrics chosen are throughput, average end-t o-end del ay, packet loss and average jitter. Simulation has been carried out for two different cases (number of act ive sessions i.e. sources either 3 or 6). W hen number of sessi ons and dat a rates is less, all th e th ree In ternet g ateway d iscoveries h ad a significant advantage, providing hi gher and m ore st able throughput, lower packet loss and end-to-end delay. But as the number of sessi ons and dat a rat es i ncrease, i t l owers throughput because m ore links break occur due to congestion and buffer overflow. Average end-to end delay and packet loss also increases as traffic increase in all the three discoveries. However, CBR packet jitter decreases as traffic in the network increases.

Lei et al [37] gave a proactive approach, a method for integrating t he Ad Hoc rout ing prot ocol wi th Mobile IP routing Pro tocol. Th is in tegration resu lts in a combined route table. Routing within Ad Hoc domain is provided by routed, a m odified versi on of RIP (routing Information Protocol), which is implemented on each m obile node. This integration enables foreign agents to participate in the Ad Hoc network routing.

In [33] t he aut hor gave an analytical analysis on Internet gateway discovery algorithms. Routing overheads for t he proact ive and react ive and hybrid methods are derived. Authors also proposed an “adapt ive gat eway advertisement” with a dynamically adjustable TTL with an optimum value of two. However, the authors did not show the performance with varying interval times and additional traffic within the MANET cluster and do not give results of gateway discovery and handover times

Hossam El-Moshrify et al. [6] proposed a solution in which mobile nodes can access the Internet via a stationary gateway node or access point. Three proposed approaches for Internet gateway discovery are implemented and investigated. Also, the effect of the mobile terminals speed and the number of gateways on the network performance are studied and com pared. A mobile


25

node uses no l oad bal ancing approach t o effi ciently discover an Internet gateway in this proposal.

Rafi U Zaman et al. [38] proposed t wo gateway load balancing strategies for Int egration of Int ernet and MANET whi ch are based on l oad bal anced rout ing protocols called WLB-AODV and Modified-AODV. The proposed st rategies have been si mulated using ns-2 simulator. Their simulation resu lts sh ow th at th e strateg y based on WLB-AODV performs better than the one based on Modified-AODV.

6.2 Mobile Internet Gateway Approaches

In [36] , Am mari et al . proposed a mobile gateway based on t hree-layer approach usi ng bot h M obile IP protocol and DSDV Ad Hoc routing protocol (Fig. 7). The first l ayer cont ains M obile IP foreign agents; the second layer includes mobile gateways and mobile Internet nodes, which are one-hop away from M obile IP forei gn agents; the third layer has all MANET nodes and visiting mobile Internet nodes that are at least one-hop away from mobile gateways. The second l ayer i s t o provi de Int ernet connectivity to MANET nodes and, t hus to help establish interaction bet ween M ANET nodes and t he Int ernet. Mobile gateways are powerful M ANET nodes and are designed in a way t o use bot h M obile IP prot ocol when they communicate with the In ternet. The DSDV protocol is u sed fo r ro uting with in th e MANET. Th e integration framework considers using some border MANET nodes to connect the rest of MANET nodes t o t he Int ernet. These MANET nodes are referred as mobile gateways. A mobile gateway selects a closest and/or a least loaded foreign agent based on the distance a nd the load criteria. MANET nodes select a closest and/or least loaded mobile gateway.

Mobile IPForeign Agents

Mobile Gateways & Mobile Internet Nodes

MANET Nodes & Visiting Internet Nodes

Mobile IP Foreign Agents

MGMG

MG

Fig 7. Three layered mobile gateway based architecture

Khan et al . [11] proposed a new approach for integrating MANET with the Int ernet by devi sing a protocol nam ed Efficient DSDV (Eff-DSDV). The proposed framework uses one of the Ad Hoc mobile nodes as a Mobile Internet Gateway (MIG), which acts as a bridge between the two networks. The M IG runs t he Eff-DSDV protocol and takes care of t he addressi ng mechanisms to ensure the transfer of packets between MANET and Internet. Thi s st rategy does not requi re t he flooding of the gateway advertisements for regi stration of mobile nodes wi th M IG. Ad Hoc rout ing prot ocol EFF-DSDV and Mobile IP coordinate with each other to provide the connectivity . Eff-DSDV follows the conventional DSDV, however i t reduces t he packet l oss due to broken links.

7. Summary of Current Proposals

Table 1 sum marizes som e of the m ain features of proposed architectures.

Table 1: Summary of Features of Existing Proposals

Proposed Architecture/ Protocol

Gateway Discovery Approach

Mobile IP Support

Ad Hoc Routing Protocol

Multiple Gateways support

Bin et al. [23] Reactive YES AODV YES

Ratanchandani et al. [24] Hy brid YES AODV YES

Hamidian et al. [27] Proactive, Reactive, Hybrid NO AODV YES

Rosenschon [7] Proactive --- AODV YES

Sun et al. [25] Proactive, Reactive YES AODV YES


26

Broch et al. [26] Reactive YES DSR NO

Jonsson et al. [21] Proactive, Reactive YES AODV YES

Lee et al. [30] Hybrid (RMD based) YES DSR YES

Tseng et al. [31] Proactive, Reactive YES DSDV YES

Xie et al. [40] Proactive YES DSDV NO

R. Wakikawa [2,28] Proactive, Reactive IPv6 Generic YES

E.M. Belding-Royer [29,18] Proactive YES AODV NO

C. Jelger et al. [32] Proactive approach - - YES

Ghassemian [34] Proactive, Reactive - AODV YES

Rakesh Kumar et al. [8] Proactive, Reactive, Hybrid NO AODV YES

Lei et al [37] Proactive YES Modified version of RIP YES

Ruiz [33] Hybrid (MBC) NO AODV NO

Hossam El-Moshrify et al. [6]

Proactive, Reactive, Hybrid NO AODV YES

Rafi U Zaman et al. [38] Reactive - AODV YES

Ammari et al. [36] Reactive YES DSDV YES

Khan et al. [11] Proactive YES DSDV YES

8. Conclusions and Future Works

In this paper, we analyzed In ternet co nnectivity of MANETs via fi xed and m obile Int ernet gat eways and pointed out limitations in t he exi sting approaches. It provides a good insight t o t he research com munity for further m odification and revi ew. M obile nodes can connect to Internet gat eways of di fferent t ypes. Several approaches have been proposed for integrating MANETs with Internet in recent years. Fixed and mobile gateways are u sed to ach ieve th e in tegration task . So me existing proposals do not consider m ultiple gateways and hence lack mechanisms for load balancing and scalability. We have report ed t echnical sol utions proposed for Internet connectivity o f MANETs v ia In ternet g ateways. Th e proactive gat eway di scovery consum es excessive network resources due t o the frequent advertisement flooding. Though, m any researchers have proposed proactive gateway discovery so lutions, it lack s lo ad balancing, hence there is a need t o propose a gat eway discovery prot ocol t hat t akes into consideration about path lo ad. Th e so licitation is a b roadcast, wh ich consumes network resources, just as t he fl ooding of advertisement does. C urrently, no exi sting schem e ever

considers the disadvantages of broadcasting solicitations. In case of hy brid gat eway di scovery approaches, a significant di sadvantage i s t hat i n t he proposed approaches some roaming nodes beyond the TTL scope of the Internet gateway m ay never receive gateway advertisements. Moreover, some mobile nodes that have registered with the gateway can not receive the latest advertisement broadcast to update registrations with their home agents.

There are m any challenges in th e in tegration o f MANETs and the Internet such as mismatches regarding their infrastructure, topology and m obility m anagement mechanisms. Based o n th e resu lts o f th is work, we believe th at an in teresting fu ture research to pic is th e work on adaptive and mobile Internet gateway discovery mechanism. In ad dition to adaptive gat eway di scovery and address autoconfiguration, there are other areas related to interworking with fi xed net works i n whi ch there is still a lot to do. These areas include among others, sel ection of an optimal Internet gateway, improved DAD (Duplicate Address Detection) mechanisms, effi cient support of DNS, di scovery of application and net work servi ces, network authentication, an integrated security m echanisms and providing Qu ality o f Serv ice (Qo S), load balancing,


27

avoiding dead zones and seam less roam ing, pricing/billing for such combined network.

In this paper we have gi ven t he cri tical revi ew of

various key techniques existing so far for M ANET-Internet in terconnectivity an d In ternet g ateway discovery. The gaps i n t he exi sting works have been identified and highlighted as and where needed.

References [1] Azzedine Boukerche, Algorithm s and Protocols for

Wireless and Mobile Ad Hoc Networks, Wiley–IEEE Press, November 2008.

[2] R. Wakikawa, J. Malinen, C. Perkins, A. Nilsson, and A. Tuominen, “ Internet Connectivity for Mobile Ad Hoc Networks,” Internet-Draft, dr aft-wakikawa-manet-global6-02.txt, Nov. 2002, Work in progress.

[3] R. Droms, J. Bound, B. Voltz , T. Lemon, C. Perkins, and M. Carney , “Dy namic Host Configuration Protocol for IPv6 (DHCPv6)”, Internet Draft, November 2002.

[4] C. Perkins, and P. Bhagwa t, “Highly Dynamic Destination Sequenced Distance Vector (DSDV) Routing for Mobile Computers,” ACM SIGCOMM Symposium on Communications, Architectures and Protocols, September 1994.

[5] Ra kesh Kuma r, Ma noj Misra, Anil K. S arje, “ Internet Access in Mobile Ad Hoc Networks,” International Conference on Advance Computing and Communication (ICACC 2007), pp. 759-762, February 9-10, 2007.

[6] Hossam El-Moshrify , M. A. Mangoud, M.Rizk, ”Gateway Discovery in Ad Hoc On-Demand Distance Vector (AODV) Routing for Internet Connectivity ,” 24th National Radio Science Conference (NRSC 2007), March 13-15, 2007,Faculty of Engineering, Alexandria University Alexandria 21544., Egypt.

[7] M. Rosenschon, T. Manz, J. Habermann, V. Rakocevic,” Gateway Discovery Algorithm fo r Ad-Hoc networks using HELLO Messages,” International W orkshop on W ireless Ad Hoc Networks IWWAN 2005, London, May 2005.

[8] Rakesh Kumar, Manoj Misra, Anil K. Sarje, “ A Simulation Analysis of Gateway Discove ry for Internet Access in Mobile Ad Hoc Networks,” International Journal of Information Processing (ISSN 0973-8215) Vol. 2, Number 2, pp. 52-64, 2008.

[9] C. E. Pe rkins, E. M. Roy er, a nd S. R. Da s, “IP Address Autoconfiguration for Ad Hoc Networks,” Internet Draft, Nov. 2001, Work in progress.

[10] Kurose J. F, Ross K. W., Computer Networking: A Top Down Approach Featuring The Internet, 3rd edition, Addison Wesley (2003).

[11] Khaleel Ur Rahman Khan, A V Reddy and Rafi U Zaman, “An Efficient Integrated Routing Protocol for Interconnecting Mobile Ad Hoc Network and the Internet,” International Journal of Computer and Electrical Engineering, vol. 1, No. 1, pp. 32-39, April 2009.

[12] Network WORKING group; Routing Information Protocol; http://www.gnu.org/copyleft/gpl.html

[13] Network Working Group; OSPF Version 2; http://www.ietf.org/rfc/rfc1247.txt

[14] J. D. Solomon, “Mobile IP: The Internet Unplugged,” Prentice-Hall PTR, 1998.

[15] Network Working Group; A Border Gateway Protocol (BGP); http://www.ietf.org/rfc/rfc1058.txt?number=1058.

[16] H.W. Cha, J.S. Park, and H.J. Kim, “Extended Support for Global Connectivity for IPv6 Mobile Ad Hoc Networks,” Internet-Draft “draft-c ha-manet-extended-support-globalv6-00.txt” October 2003.

[17] S. Thomson, and T. Narten , ”IPv6 Stateless Address Auto configuration,” IETF RFC 2462, December 1998.

[18] E.M. Belding-Roy er, Y. Sun, and C. Perkins, “Global Connectivity for IPv4 Mobile Ad Hoc Networks, IETF Internet Draft, Nov. 2001, Work in progress.

[19] K. Weinger, “Passive Duplicate Address Detection in Mobile Ad Hoc Networks,” University of Karlsruhe Germany, IEEE WCNC 2003, March 2003.

[20] C. E. Perkins, and E. M. Belding-Royer, “Ad Hoc On-Demand Distance Vector (AODV) Routing,” draft-perkins-manet-aodvbis-00.txt, Internet Draft, 19 October 2003.

[21] U. Jonsson, F. Alriksson, T. Larsson, P. Johansson, and G.M. Ma quire, “MIPMANET: M obile IP for Mobile Ad Hoc Networks,” Proceedings of IEEE/ACM W orkshop on Mobile and Ad Hoc Networking and Computing (MobiHoc 2000), Boston, MA USA, pp. 75-80, August 1999.

[22] Y. Sun, E.M. Belding_Royer, C.E. Perkins,” Internet Connectivity for Ad Hoc Mobile Networks,” International Journal of Wireless information Networks, Special Issue on Mobile Ad Hoc networks (MANETs): Standards, Research, Applications 9 (2) (2002) 75-88.

[23] S. Bin, S. Bingxin, L. Bo, H. Zhonggong, and Z. Li, “Adaptive Gateway Discovery Scheme for Connecting Mobile Ad Hoc Networks to the Internet,” P roceedings of International Conference on Wireless Communications, Networking and Mobile Com puting, vol. 2, pp. 795-799, 2005.

[24] P. Ratanchandani, and R. Kravets, “A Hybrid Approach to Internet Connectivity for Mobile Ad Hoc Networks,” in Proceedings of the IEEE WCNC 2003, New Orleans, USA, vol. 3, pp. 1522-1527, March 2003.

[25] Y. Sun, E. M. Belding-Roy er, and C. E. Perkins, “Internet Connectivity for Ad Hoc Mobile Networks,” International Journal of Wireless Information Networks, Special Issue on Mobile Ad Hoc Networks (MANETs): Standards, Research, Application. 9(2), pp. 75-88, April 2002.

[26] J. Broch, D.A. Maltz , and D.B. Johnson, “Supporting Hierarchy and Heterogeneous Interfaces in Multi-Hop Wireless Ad Hoc Networks ,” in P roceedings of the IEEE International Sy mposium on Parallel Architectures, Algorithms, and Networks, Pert h, Western Australia, pp. 370-375, June 1999.

[27] A. Hamidian, U. Korner, and A. Nilsson, “A Study of Internet Connectivity for Mobile Ad Hoc Networks in NS2”, Department of Co mmunication Sy stems, Lund Institute of Technology, Lund University, January 2003.

[28] Wakikawa R., Malinen J., Perkins C., Nilsson A., “ Global Connectivity for IPv6 Mobile Ad Hoc Networks” In IETF Internet Draft 2003. http://www.wakikawa.net/Research/paper/draft/manet/draft-wakikawa-manet-globalv6-03.txt


28

[29] E.M. Belding-Roy er, Y. Sun, C.E. Perkins, “Global Connectivity for IPv4 Mobile Ad Hoc Network, IETF Internet-Draft, draft-ietf-man et-globalv4-00.txt, November 2001.

[30] J. Lee, D. Kim, J. J. Garcia-Luna-Aceves, Y Choi, J Choi, and S. Nam, “Hybrid Gateway Advertisement Scheme for Connecting Mobile Ad Hoc Networks to the Internet, “ in Proceedings of the 57th IEEE VTC2003, Jeju, Korea, vol. 1, pp. 191-195, 2003.

[31] Y. C.Tseng, C. C. Shen, a nd W. T. Chen , “Mobile IP and Ad Hoc Networks: An Integration and Implementation Experience,” Technical Report, Department of Computer Science and Information Engineering, Chaio Tung University, Hsinchu, Taiwan, 2003.

[32] C. Jelger , T. Noel, A. F rey, “Gateway and Address Auto configuration for IPv6 Ad Hoc Networks,” IETF internet-Draft, draft-jelger-m anet-gateway-autoconf-v6-02.txt, April 2004

[33] P. Ruiz, A. Gom ez-Skarmeta, “ Enhanced Internet Connectivity for hybrid Ad Hoc networks through adaptive gateway discovery ,” International Conference on Local Computer Networks, LCN 04, Tampa, Florida, November 2004.

[34] M. Ghassemian, P. Hofmann, C. Prehofer, V. Friderikos and H. Aghvami, “ Performance Analy sis of Internet Gateway Discovery Protocols in Ad Hoc Networks,” IEEE WCNC 2004, Atlanta, Georgia, USA.

[35] D. Johnson, D. Maltz, J. Jetcheva, “ The Dy namic Source Routing Protocol for Mobile Ad Hoc Networks, “ Internet Draft, draft-ietf-manet-dsr-07.txt, work in progress, 2002.

[36] H. Ammari and H. El-Rewin i,” Integration of Mobile Ad Hoc Networks and the Internet using Mobile Gateway s,” Proceeding of the 4th International W orkshop on Algorithms for Wireless, Mobile, Ad Hoc and Sensor Networks (WMAN04), Santa Fee, New Mexico, USA, April 26-30, 2004.

[37] Lei H., Perkins C., “Ad Hoc Networking with Mobile IP,” in Proceedings of 2nd European Personal Mobile Communication Conference. (1997).

[38] Rafi U Zam an, Khaleel Ur Rahm an Khan, M .A.Razzaq and A. Venugopal Reddy, “A Simulation Based Comparison of Gateway Load Balancing Strategies in Integrated Internet-MANET”, 17th International Conference on Advanced Computing and Communication (ADCOM’09) 14-17 Decem ber 2009. IISC Bangalore, Published by IEEE Computer Society, pages: 270 – 272.

[39] E.M. Roy er, C-K. Toh, “A Review of Current Routing Protocols for Ad Hoc Mobile W ireless Networks,” IEEE Personal Communications Ma gazine, April 1999, pp. 46-55.

[40] B. Xie, A. Kumar, “ Integrated connectivity framework for Internet and Ad hoc Networks,” First International Conference on Mobile Ad hoc and Sensor Sy stems, October 2004, Fort Lauderdale, Florida, USA.

Rakesh Kumar received his B.E. degree in Computer Engineering from M.M.M. Engineering College Gorakhpur (U.P.), India in 1990 and his M.E. In Computer Engineering from S.G.S. Institute of Technology and Science, Indore, India in 1994. Since January 2005, he has been a PhD student in the Department of Electronics & Computer Engineering at Indian Institute of Technology, Roorkee, India. He is a life member of CSI, ISTE and also a Fellow of IETE. His main research interests lie in Mobile Ad Hoc Routing, Quality of Service Provisioning, MANET-Internet Integration and Performance Evaluation.

Anil K. Sarje is Professor in the department of Electronics & Computer Engineering at Indian Institute of Technology Roorkee, India. He received his B.E., M.E. and PhD degrees from Indian Institute of Science, Bangalore in 1970, 1972 and 1976 respectively. He served as Lecturer at Birla Institute of Technology & Science, Pilani, for a short period before joining University of Roorkee (now Indian Institute of Technology Roorkee) in 1987. Dr. Sarje has supervised a large number of M.Tech. dissertations and guided several Ph.D. theses. He has published a large number of research papers in the International and National journals and conferences. He has also served as referee for many reputed Journals like IEE Proceedings, IEEE Transaction on Reliability, Electronics Letters, etc. He has been on a number of AICTE and DOEACC committees. He was a member of All India Board of Information Technology during years 2000-2003. He is a senior member of the Institute of Electrical and Electronics Engineers (IEEE). His research interests include Distributed Systems, Computer Networks, Real Time Systems and Network Security.

Manoj Misra is Professor in the department of Electronics & Computer Engineering at Indian Institute of Technology Roorkee, India. He received his B.Tech. degree in 1983 from H.B.T.I., Kanpur and M.Tech. from University of Roorkee in 1986. He did his Ph.D. from Newcastle upon Tyne, UK and joined Electronics & Computer Engineering Department, University of Roorkee (now Indian Institute of Technology Roorkee) in August 1998 as Assistant Professor. Before joining University of Roorkee, he worked in DCM, CMC Ltd., New Delhi, H.A.L. Kanpur and H.B.T.I. Kanpur. He has completed an AICTE funded project "A CORBA framework for distributed mobile applications", as a co-Investigator with Dr. R. C. Joshi. Dr. Misra has supervised a large number of M.Tech. Dissertations and guided several Ph.D. Theses. He has published a large number of research papers in International and National journals and conferences. He is a member of the Institute of Electrical and Electronics Engineers (IEEE). His research interests include Distributed Computing and Performance Evaluation.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 4, No 6, July 2010 ISSN (Online): 1694‐0784 ISSN (Print): 1694‐0814 29

A New Approach to Supervise Security in Social Network through Quantum Cryptography and Non-

Linear Dimension Reduction Techniques

Lokesh Jain1 and Prof. Satbir Jain2

1 M.Tech. Final Year, NSIT, Dwarka, New Delhi, INDIA

2 Department of Computer Science, NSIT, Dwarka, New Delhi, INDIA

Abstract

Social networking sites such as Orkut, Tribe, o r Facebook allow millions of individuals to create online profiles and share personal information with vast networks of friends - and, often , unknown numbers of strangers. Some of the information revealed inside these networks is private and it is possible that corporations could use learning algorithms on the releas ed data to predict undis closed private information. To find the patterns of information revelation and the ir s ecurity im plications, we an alyze the onlin e behavior of wiki-vote data set and evaluate the amount of information th ey disclose and s tudy their dimension used for reduc tion. I n this p aper we conclude that dimension reduction is one of the factors through we can achieve the security and maintain the integrity of dataset. We highlight various Non-Linear dimensio n reduction tech niques with quantum cr yptography to prod uce the desir e result and show the com parative res ult with line ar dimension reduction technique. Keywords: Social network, Security, Quantum cryptography, Non-Linear Dimension reduction

1. Introduction Recently, online s ocial network has em erged a s a promising are a wi th m any p roducts an d a huge number of users . W ith the development of information retrieval and search engine techniques, it becomes very conve nient to extract us ers’ pers onal information that is read ily available in various social networks. Malicious or curious users take advantage of t hese t echniques t o collect ot hers’ p rivate information. Therefore, it is critical to enable users to control t heir i nformation di sclosure and effectively maintain security over online social networks. One of the challenges in so cial network is security. Although security preservation in data publishing has been st udied ext ensively an d sev eral im portant

models such as k-anonymity and l-diversity as well as many ef ficient algorithms have been proposed, most of th e ex isting stu dies can deal with relati onal d ata only. T hose methods ca nnot be a pplied t o soci al network data st raightforwardly. Security m ay be break i f a s ocial net work i s rel eased i mproperly t o public. In p ractice, we need a systematic method to anonymize social network da ta before it is released. However, an onymizing soci al net work d ata i s much more ch allenging th an anony mizing r elational data on which most of the previous work focuses. One of t he ways to achieve the security in Social network is to red uce the di mension of w hole social network by linear and non-linear dimension reduction technique. The dimension of the data, is the number of variables that are measured on e ach observation.the pr oblems with high-dimensional datasets is th at, in many cases, not all th e measured variables are “im portant” for understanding t he underlying p henomena of interest. Wh ile certain computationally expensi ve no vel methods can construct predictive models with high accuracy from high-dimensional d ata, it is still o f in terest in many applications t o red uce t he dimension of t he ori ginal data prior to any modeling of the data. By reduce the dimension of these m odels we can als o a chieve the security of the original data set. So in this paper, in the second section we introduce the soci al net work a nalysis. In t he t hird a nd fourth section we refers t o t he no n-linear dimension reduction t echniques a nd i n ne xt t wo se ction we represents our design approach a nd experim ental result and in the last section we concluded with future work and conclusion.


2. Social Network Analysis Social net work analysis [SNA] i s t he m apping an d measuring of relationships and flows between people, groups, organizations, computers, UR Ls, and ot her connected information/knowledge entities. The nodes in th e network are th e p eople and groups while the links show relationships or f lows between the nodes (Fig.1). SNA pr ovides both a visual and a mathematical anal ysis o f hum an rel ationships. Management cons ultants use t his methodology wi th their business clien ts and call it Organ izational Network An alysis [ ONA]. To un derstand n etworks and th eir participants, we ev aluate t he location of actors in the network. M easuring t he net work location is find ing th e centrality of a node. These measures give us i nsight i nto t he various roles a nd groupings in a

Figure 1: social network structure

network - who are t he co nnectors, m avens, leaders , bridges, isolates, where are t he clusters and who is in them, who i s in the core of the network, and who is on the periphery? So i n nu t sh ell w e can say t hat N etwork an alysis is the study of social relations among a set of actors. It is a field of study -- a set of phenomena or data which we seek to understand. In t he process o f working in th is field, n etwork researchers h ave de veloped a set of di stinctive theoretical p erspectives as well. So me o f th e hallmarks of these perspectives are:

Focus on relationships between actors rather than attributes of actors.

Sense of int erdependence: a molecular rather atomistic view.

Structure affects substantive outcomes. Emergent effects.

Network th eory is sy mpathetic with systems th eory and co mplexity th eory. A social n etwork is also characterized by a distinctive methodology encompassing techniques for collec ting data, statistical analysis, visual representation, etc.

3. Reason for choosing Non-Linear Dimension Reduction Technique In case of s ocial network, t he size of the data set is large an d dat a set has vari ous di mensions. Due t o severity of so cial network data, it is v ery difficult to secure the dat a. Consi der a dataset re presented as a matrix (or a database tab le), su ch th at each row represents a set of attributes (or fe atures or dimensions) t hat desc ribe a part icular i nstance of something. If t he number of attributes is larg e, th en the space o f unique p ossible ro ws i s e xponentially large. Thu s, t he larg er t he d imensionality, th e m ore difficult it bec omes to sa mple the space. T his causes many problems. Al gorithms that ope rate on high-dimensional d ata t end t o have a very high t ime complexity. M any m achine learning al gorithms, fo r example, strugg le with h igh-dimensional data. Th is has b ecome k nown as th e cu rse of dimensionality. Reducing data i nto fewe r d imensions o ften m akes analysis algorithms more efficient, and can help machine learning al gorithms m ake more accurate predictions. By reduci ng t he di mension of dat aset, we can al so achieve t he s ecurity. Because dim ension reduction only represents the a bstract feature of a particular data set. Data ab straction sho ws only th ose featu res that are essen tial to represent th e d ata an d h ide t he remaining details. Hence data hiding is a one way to achieve th e secu rity. In th is pape r we ana lyze that non-linear dimension reduction a m ore efficient way as com pared to linear D imension r eduction techniques. These techniques not only use for feature selection and extraction but can also be used for security purposes. 4. Introduction to Non-Linear Dimension Reduction Techniques Advances in data co llection an d st orage cap abilities during t he past decade s have led to a n inform ation overload i n m ost sci ences. Researchers w orking i n domains as diverse as e ngineering, a stronomy, biology, rem ote sensi ng, eco nomics, and c onsumer transactions, face large r a nd larger observations a nd simulations on a daily basis. Such datasets, in con trast with sm aller, m ore traditional datasets that have been studied extensively in th e p ast, present new challen ges i n data an alysis. Traditional statistical methods b reak d own p artly because of the increa se in the number of observations, but m ostly because of t he increase i n the num ber of va riables associated with each


observation. The dimension of the data is the number of variables that are measured on each observation. We su bdivide tech niques for dimensionality reduction i nto con vex an d non-convex t echniques (Fig. 2) . C onvex t echniques opt imize an ob jective function t hat doe s not c ontain any l ocal opt ima, whereas n on-convex techniques op timize objective functions t hat do c ontain l ocal opt ima. The fu rther subdivisions in th e tax onomy ar e d iscussed in th e review in the following two sections:

Figure 2: Taxonomy of Dim. Reduction techniques

4.1 Convex Techniques for Dimensionality Reduction Convex techniq ues fo r d imensionality redu ction optimize an objective functi on that does not contain any local optima, i.e., t he s olution s pace is conve x. Most of t he selected dimensionality redu ction techniques fall in the class of convex techniques . In these techn iques, th e obj ective function usually h as the form of a (ge neralized) Rayleig h q uotient: th e

objective function is of the form Ø(Y)= . It is well kn own t hat a function of th is form can b e optimized by sol ving a ge neralized ei genproblem. One technique (Maxim um Va riance Unfolding) solves a n ad ditional sem idefinite pro gram using a n interior po int method. We su bdivide co nvex dimensionality redu ction tech niques in to tech niques that perform an eigen decomposition of a fu ll matrix and t hose t hat pe rform an e igendecomposition of a sparse matrix. 4.1.1 Full Spectral Techniques Full spectral techniques for dimensionality reduction perform an ei gendecomposition of a full matrix that captures the covariances bet ween dim ensions or the pairwise similarities b etween datapoints (possibly in a feature s pace that is cons tructed by means of a kernel f unction). I n t his su bsection, we have five such techniques: (1) PC A / classical scaling, (2)

Isomap, ( 3) Kernel PCA , (4) M aximum Varia nce Unfolding, and (5) diffusion maps. 4.1.2 Sparse Spectral Techniques In t he p revious subsection, we di scussed fi ve techniques t hat co nstruct a l ow-dimensional representation of th e high-dimensional d ata by performing an eig endecomposition of a full matrix. In co ntrast, t he fo ur t echniques discussed i n t his subsection solve a sparse (generalized) eigenproblem. All prese nted sparse spect ral t echniques only foc us on retaining l ocal st ructure of t he dat a. We di scuss the sp arse sp ectral di mensionality red uction techniques (1) LLE, (2) La placian Eige nmaps, (3) Hessian LLE, and (4) LTSA 4.2 Non-convex Techniques for Dim ensionality Reduction We ha ve a no n-convex t echniques f or multidimensional scaling that forms an alternative to classical scaling called Sa mmon m apping , a technique based on trainin g m ultilayer n eural networks ,and t wo t echniques t hat c onstruct a mixture of local l inear models and pe rform a gl obal alignment of t hese l inear m odels. S o we ha ve following n on-convex t echniques (1 ) Sam mon Mapping, (2 ) Mu ltilayerAutoencoder, (3) Lo cally Linear Coordination (LLC ) and (4 ) M anifold Charting. 5. Introduction to Quantum cryptography Quantum cryptography was proposed by Bennett and Brassard i n 1984, who al so defi ned t he f irst QK D protocol, called BB84. At time o f writing, a h andful of research tea ms around the world have s ucceeded in b uilding and operating quantum cry ptographic devices. Funda mental aspects of quantum physics – unitarity, the uncertainty princip le, and the Einstein-Podolsky-Rosen io lation of Bell’s in equalities – suggest a new pa radigm fo r key di stribution: quantum cryp tography. In itial ex periments seem to confirm th e u tility o f th is paradigm. Assumin g th at the th eoretical m odels con tinue to b e confirm ed in the use of actual de vices, t he fundam ental laws of nature can be invoked to assure the confidentiality of transmitted data. Qu antum cryp tography – m ore properly termed Qu antum Key Distribu tion, QKD – employs t wo di stinct c hannels. One i s use d for transmission o f q uantum key material by very di m (single p hoton) l ight p ulses. The other, p ublic channel carries all message traffic, incl uding the cryptographic pr otocols, encrypted u ser t raffic,


etc.QKD co nsists of t he t ransmission of raw key material, e.g., as di m pul ses of l ight f rom Al ice t o Bob, via the quantum channel, plus processing of this raw material to derive the actual keys (Fig. 3). This processing i nvolves p ublic com munication (key agreement pr otocols) between Alice and Bob, conducted in the public channel, along with specialized QKD algorithms. The resu lting keys can then be used fo r c ryptographic purposes, e. g., t o protect user t raffic. By the laws o f quantum physics, any eavesdropper (Eve) that snoops on the quantum channel will cau se a m easurable d isturbance to th e flow of single photons. Alice and Bob can detect this, take appropria te steps in re sponse, and hence foil Eve’s attempt at eavesdropping.

Figure 3: Quantum - Key Distribution

6. Design Approach For achieve security in social network, we proposed a model usi ng qua ntum cry ptography t o achi eve security (Fi g. 4). In th is m odel first we co llect th e information regar ding t he s ocial net work dat a and after that reduce the dimension of t he data set usi ng various dimension re duction t echnique. Aft er reducing th e data, we co vert th e d ata in to digitized form and to encrypt the data we use the MD5 (Message Di gest 5) t echnique. Generally, in stead of MD5, we ca n al so use any ot her e ncryption techniques, such as DSA, AES and s o on, to decrypt the message. For t he secure key distribution purpose we ca n a pply t he quantum cry ptography. Quantum cryptography provides a secure way to distribute the key on the basis o f qu antum th eory. Generally, it is very di fficult t o im plement t his key dist ribution technique because i t req uires Ph oton l ight p ulses (PLP) t hrough whi ch p hoton t ravel fr om one use r side t o a nother si de. So, in t his pa per we j ust proposed t he QKD ( Quantum Key Di stribution) f or key distribution. After encrypt the key we transfer the

data thr ough SNSs site and at th e other end, user decrypt the data in the reverse order.

Figure 4: Design Approach to preserve security

Hence, i n t his pa per we proposed a secure a nd authenticate techni que for s ocial network data sets which is more secure as com pared to earlier techniques. 7. Experimental Result In our experiments on ‘wiki vote’ datasets, we apply the ten techniques for dimensionality reduction on the high-dimensional re presentation o f t he dat a. Subsequently, we assess th e quality of th e resulting low-dimensional d ata represen tations by ev aluating to wh at ex tent th e lo cal st ructure of th e d ata is retained. Th e eval uation i s p erformed by measuring the g eneralization er rors o f 1-n earest n eighbor classifiers that are trained on the l ow-dimensional data representation. Wiki-is a free en cyclopedia written co llaboratively by vol unteers aro und t he world. A sm all part of Wikipedia co ntributors a re adm inistrators, who are users with access to add itional technical features that aid in m aintenance. In order for a use r to become an administrator a R equest for adm inship (RfA) is issued a nd t he Wikipedia c ommunity vi a a public discussion or a vote deci des w ho t o promote t o adminship. U sing t he l atest com plete dump of Wikipedia pa ge edit hi story (f rom January 3 2 008) we extracte d all ad ministrator elections and vote history dat a. Thi s ga ve us 2,794 elections wi th 103,663 total v otes and 7 ,066 u sers p articipating in the elections (either casting a vote or being voted on). Out of these 1,235 elections resulted in a s uccessful promotion, while 1,559 elections did not result in the promotion. About half of the votes in the dataset are by existing admins, while the other half comes from ordinary Wikipedia users (Table 1).


The network contains all th e users and discussion from the in ception of Wikipedia till Jan uary 2 008. Nodes in the network represent wikipedia users and a directed e dge from node i to nod e j represent that user i voted on user j.

Table 1: Representation of Wiki-Vote Data set

Nodes 7 115 Edges 1 03689 Nodes in largest WCC 7066 (0.993) Edges in largest WCC 103663 (1.000) Nodes in largest SCC 1300 (0.183) Edges in largest SCC 39456 (0.381) Average clustering coefficient 0.2089 Number of triangles 608389 Fraction of closed triangles 0.1255 Diameter (longest shortest path) 7 90-percentile effective diameter 3.8 We per form the No n-linear di mension red uction techniques on t he wi ki-vote dat a set an d get t he following result:

Original data set

PCA

Isomap

Autoencoder

Diffusion Map

Hessian LLE

Kernal PCA


Sammon Mapping

Laplacian Eiganmap

LLE

LLC Man. charting

In th e ab ove d iagram w e r epresent th e g raphical representation of the dimension reduction of various Non-Linear techniques. After reduce the di mension of dat a, we convert the reduce d ata into d igitized form by sampling technique and encrypt the data using MD5 technique. Hence, we ca n fi nd a way t o achi eve s ecurity i n social network.

Now, we rep resent th e com putational co mplexity, memory and running time of these techni ques in the table 2 shown below:

Table 2: Measures of Dimension reduction techniques

8. Conclusion and Future work In this paper we proposed a more secure, reliable and authenticate quant um crypt ography base d technique followed b y co nvex and n on-convex dimension reduction techn ique. Gen erally i t is v ery d ifficult to implement t o qua ntum cry ptography due t o hi gh overhead of radiation problem of p hoton a s wel l as physical i mplementation o f Ph oton Li ght Pul se (PLP). S o i n fut ure we c an i mplement pr oposed model wi th Q DK an d use more efficient non-linea r dimension reduction technique whose complexity and running ti me i s lesser th an t he presen t tech niques. This technique can also introduce a m ore secure and authenticate comm unication channel fo r co mmunity and blog creation. 9. References [1] W.E. Arnoldi. The princip le of minimized iter ation in

the solution o f the m atrix eigenv alue pro blem. Quarterly of Applied Mathematics, 9:17–25, 1951.

[2] Bar-Hillel, T . Hertz, N. S hental, and D. W einshall. Learning a Mahalanob is me tric from equivalence constraints. J ournal of M achine L earning Res earch, 6(1):937–965, 2006.

[3] S.P. Bo yd and L. Van denberghe. C onvex optimization. Cambridge University Press, New York, NY, USA, 2004.

S. No

Technique C omputational Complexity

Running time(sec.)

Memory

1 PCA O(D3) 2. 41 O(D2) 2 Sam mon

Mapping O(n3) 3 0.51 O(n3)

3 I somap O(n3) 2 2.40 O(n2) 4 K ernal

PCA O(n3) 2 7.45 O(n2)

5 D iffusion Map

O(n3) 2 5.77 O(n2)

6 A uto Encoders

O(inw) 67 5.29 O(w)

7 LLE O(pn2) 12 .62 O(pn2) 8 Laplacian

Eigenmap O(pn2) 10 .47 O(pn2)

9 Hessian LLE

O(pn2) 16 .33 O(pn2)

10 LLC Mani.

Charting

O(imd3) 10 .09 O(nmd)


[4] E.W. Dijkstra. A note on two problems in con nexion with graphs. Numerische Mathematik, 1:26 9–271, 1959.

[5] P. Demartines and J. H´ erault. Curvilinear component analysis: A self-organ izing neural network for nonlinear mapping of data sets. IEEE Transactions on Neural Networks, 8(1):148–154, 1997.

[6] D.L. Donoho and C. Grimes. Hessian eigenmaps: New locally lin ear embedding techniques for high-dimensional data. Proceedings of the National Academy of Sciences, 102(21):7426–7431, 2005.

[7] R. Duraiswami and V. C. Raykar. The manifolds of spatial h earing. In P roceedings of Internat ional Conference on Acous tics, S peech and Signal Processing, volume 3, pages 285–288, 2005.

[8] Faloutsos and K.-I. Lin. FastMap: A fast algorithm for indexing, datamining and visu alization of traditional and m ultimedia datasets. In Proceed ings of the 1995 ACM SIGMOD International Confer ence on Management of Data, pages 163–174, New York, NY, USA, 1995. ACM Press.

[9] K. Fukunaga. Introduction to Statistical Pattern Recognition. Academic Pre ss Professional, Inc., San Diego, CA, USA, 1990.

[10] G.E. Hinton, S. Osindero, and Y. Teh. A fast learning algorithm for d eep belief nets. Neural Computation , 18(7):1527–1554, 2006.

[11] H. Hoffmann. Kernel PC A for novelty d etection. Pattern Recognition, 40(3):863–874, 2007.

[12] J.A. Lee and M. Verleysen. Nonlinear dimensionality reduction of data manifold s with essential loops. Neurocomputing, 67:29–53, 2005.

[13] J.A. Lee and M. Verleysen. Nonlinear dimensionality reduction. Springer, New York, NY, USA, 2007.

[14] A.M. Posadas, F. Vidal, F. de Miguel, G. Alguacil, J. Pena, J. M. Iba nez, and J. Morale s. Spa tial-temporal analysis of a seismic s eries us ing the prin cipal components method. Jour nal of Geoph ysical Research, 98(B2):1923–1932, 1993.

[15] N. Gisin et al, “Quantum cryptography,” Rev . Mod. Phys., Vol. 74, No. 1, January 2002.

[16] Elliott, “Building the quantum network,” New J. Phys. 4 (July 2002) 46.

[17] G. Brassard and L. Salvail, “Secret key reconciliation by pub lic discussion,” Lect. Notes in Com puter Science 765, 410. (1994).

Mr. Lokesh Jain. is a M.Tech.(IS) final year student in Netaji Subhas Institute of Technology, Delhi, India. He has completed his graduation in 2005 from UPTU, India with honors. He has been associated with many academic institutes since last 4 years as faculty. He has been published more than 10 national and international papers on database and social networking. His area of interest is Database, Machine Learning, Discrete Mathematics and fuzzy Logic. Presently he is working on the social network security and privacy project.

Dr. Satbir Jain. has been associated with many international societies such as IEEE, CSI and IETE. He has been published more than 50 international and national papers. He is also a member of editorial committees of many technical societies. His area of interest is Database, Datamining, Data Modeling, Object Orientation, Software Engineering and Image Processing. Presently Dr. Jain is a professor in Department of Computer science in Netaji Subhas Institute of Technology, Delhi, India.

IJCSI International Journal of Computer Science Issues, Vol. 7, Issue 4, No 6, July 2010 ISSN (Online): 1694-0784 ISSN (Print): 1694-0814 36

The Grassmannian Manifold and Controllability of the Linear Time-Invariant Control Systems

S. M. Deshmukh1, Mrs. Seema S. Deshmukh2, R. D. Kanphade3, N. A. Patil4.

1Deptt. of Electronics & Tele-Com., P.R.M.I.T. & R., Badnera -444701 (INDIA).

2Deptt. of Physics, P.R.M.I.T. & R., Badnera -444701 (INDIA).

3Principal, Dhole-Patil Engineering College, Pune-412207 (INDIA)

4Deptt. of Mathematics, Sant Gajanan Maharaj Engineering College, Shegaon-444203 (INDIA)

Abstract The idea discussed here are mainly to develope some interesting relationship between the differential geometry of certain curves and the controllability of linear time-invariant (LTI) control systems without considering any matrix riccati equation. The problem based on the basic concepts of controllability is considered here. The two point boundary value problem (TPBVP) is described here as a flow in the Grassmannian manifold. Then a simple solution to determine a control function in the Grassmannian manifold is presented that transfer the system states from initial to final values and satisfies the conditions that are equivalent to the controllability of the systems.

Keywords: Linear system, Control function, controllability, Grassmannian manifold. 1. Introduction

Consider a LTI control system in the form of

x(t)=Ax(t)+Bu(t), (1)

for xRn, uRm and A, B are constant matrices. The problem of controllability of the LTI control system (1) is considered here as a state transfer problem (STP). Thus we determine a control function u(t) that transfer the system states from the initial to final values within specified time interval of t seconds. The problem of controllability is discussed in [1], as the state transfer problem. The solution of STP is given in [1], by computing the state-transition matrix. In [2], concept of controllability is considered as a STP and proposed several methods for synthesizing a control function for steering the given initial state of the system to the origin. In [3], the

solution to STP is based on relating the given system to a family of phase-variable canonical form systems and then by using the technique of two-point interpolation. The idea about differential geometry is given in [4]-[6]. It relates with the differential relation that stitches pieces of curves or surfaces together. It relates with the curves, surfaces, the functions that define them and transformation between the coordinates that can be used to specifies them. Here, the concepts of differential geometry uses as a tool for analysis of control systems. The system (1), is described here as a flow in the Grassmannian manifold. Then a control function u(t) have synthesized in the Grassmannian manifold that transfer the system states from initial to final values and satisfies the conditions that are equivalent to the controllability of the LTI control systems. Main objective here is that the control engineers should not restrict himself to any one of the tool but should be familiar with as many as possible for analysis purpose. The interesting relationship between the differential geometry of certain curves and the controllability of LTI control systems have been discussed in [7], with matrix riccati equation. But we have developed the same concepts with different approach, without considering any matrix riccati equation.

2. Brief Review of Literature

Some basic facts about the Grassmannian manifold and certain group actions are as follows. For detail, the reader should refer the exposition given by Doolin and Martin [6].

2.1. Lie group and group action:

Since the idea of a group is purely abstract algebraic idea, the definition of a group should involve only a set of element and some algebraic relations between them. A group is a set

.


of elements, like a set of matrices, any pair of which can operate together to given another element that is also in the same set. In addition to different operation of the group, it requires some conditions, that is each element of the group and its inverse should be in the same group, one element in the group should be as an identity element and finally, the operation should be associative. For example, the set of all n × n unity matrices, U(n), forms a group with the usual matrix product as the operation. The group U(n) is a manifold, the product of two unity matrices is a unitary matrix and this operation is a manifold map [6]. The group with all these properties are called lie group or continuous transformation group. Lie group is a group that is a manifold and whose group operation yields C∞ manifold function and is associative. Group Action:

The action of a group refers to members of a group operating on nonmembers. To understand the action of a group on a manifold, we consider the Lie group-G as a closed sub-group of the group of N × N invertible matrices, Gl(N). Now, we shall state the following definition from [6]. Definition: The group G is said to act on the manifold, Gp(V), if there exist an infinitely differentiable function τ : G Gp(V) Gp(V) with the following properties. i) for all m Gp(V) , τ (e, m) = m, where e is the identity element. ii) for all g, h G and m Gp(V) , τg, τ(h,m) = τ (gh,m). We say in this case that the group G acts on Gp(V) . Let (t) be a one parameter sub-group of G with the following properties: (t)(h) = (t + h), (0) = 1, the identity. Using these properties we can show that -1(t )= (-t) and (t) = A (t) (2)

where 0

lim [ ( ) (0)] / .h

A h h

The matrix A is called the infinitesimal generator of the sub-group α. If α and A are real numbers, equation (2) has the solution (t)eAt(0) = eAt , confirming the connection of above mentioned properties with exponential. The same form holds if A is matrix, generating a matrix representation of the subgroup. Clearly the solution of the differential equation (2) is, (t) = exp At. (3) 2.2. The Grassmannian manifold:

The Grassmannian manifold is the set of all p-dimensional subspaces of an n-dimensional vector space V. It is denoted by Gp(V). Every p-dimensional subspace is denoted by WG given by the linear transformation of (n-p) × p matrix G.

Let X and W be one dimensional subspaces of two dimensional vector space V, such that V is their direct sum: V = X W. By the direct sum meant that X and W have no subspaces in common except (0,0) of V, which will be referred to as the set {0}. The situation can be visualized in Fig.1. The point p in Fig.1 belongs to the subspace WG. It is specified uniquely by x + Gx, with G a real number (a 1 x 1 matrix) and x X . Every point in WG is specified similarly by some x and every point in the plane except W itself belongs to a WG for some G. W WG

Gx p

x X

Fig. 1

The p-dim subspace X of V can be represented by a unique form, WG in Gp(V) as

: ,x

W x X and G L X WG Gx

(4)

for some matrix G, if and only if X0 ∩ W = {0}. We leave out the proof of (4) and refer the reader to [5] & [6]. Let Gl(N) be the set of all N × N invertible matrices. Also, let (t) be a one-parameter sub-group of Gl(N) and Gp(V) be regarded as the Grassmannian manifold. The action of (t) on Gp(V) is given as : Gl(N) Gp(V ) Gp(V) (5) Therefore for α(t) Gl(N) and WG Gp(V), we have (α(t),WG ) α(t)(WG). (6) That meant an integral curve, x(t)=(t)WG, in Gp(V), is formed by the action of a one parameter sub-group (t) of G as shown in Fig.2. Gp(V) Gl(N)

.

WG

(t)WG

(t)

Fig. 2.


The mental picture is that the curve x(t) being traced in Gp(V) by the evolution on the continuous transformation of the initial p-dim. sub-space, WG, under the action of the group. 3. Methodology

Let Gp(V) be a Grassmannian manifold. A curve, x(t), in a manifold Gp(V), is defined to be a differentiable function,x:R→ Gp(V),whose domain, (a, b), is an open interval of R. Then, x(t) = A (x(t)), is a differential equation on Gp(V). Now, a curve, x(t), in Gp(V) satisfying the equation x(t) = A (x(t)), and such that x(a) = xa. Such a curve is also called an integral curve starting from the point xa. At this stage, consider a curve x(t), in Gp(V), formed by the action of a one-parameter sub-group (t) of G. That is, x(t) = (t)xa. By differentiating this equation and by using the equation (2), we get the equation. x(t)=Ax(t) (7) Since A is the infinitesimal generator of a Lie group (t), the determination of which corresponds to finding the solution of the differential equation (7). This construction produces an ordinary differential equation in the form of the initial value problem with the initial value xa. The vector field is generated by the matrix A and the flow, (t) = expAt on the Grassmannian manifold Gp(V). Now, if WG be the initial p-dim. subspace in the Grassmannian manifold, Gp(V), then by the group action, Gl(N) Gp(V) Gp(V). (8) Therefore for WG Gp(V) and (t) Gl(N), (α(t), WG) α(t) (WG). (9) If LTI control system (1) is in the form of x(t)=Ax(t)+Bu(t), (10) for xRn, uRm, t[0,T], with boundary conditions x(0) = x0 and x(T) = xT . Then with some efforts it is possible to describe the system (10), as a flow in the Grassmannian manifold,Gp(V), as

x(t)=Ax(t)+Bu(t), (11) for t[a,b], with boundary conditions xa=WGO and xb=WGT . A curve x(t) = α(t)WG is an integral curve satisfying the equation (11) in Gp(V), starting from the initial point xa=WGO. For a necessary and sufficient condition for the existence of a unique solution of the boundary value problem,(11),if we consider (t) in the matrix form, then from (4),we get,

10

3

( )( )

( )a

Ga

t xt W

t x

(12)

The eqn.(12) can rewrite as

0 13 1

( )( )

( ) ( ) ( )G

E tt W

t t E t

(13)

where E(t) = 1(t) xa X. .Such a representation of (t) is possible if and only if -1(t) exists. Hence the results. This is very important result, that gives condition for the existence of a unique solution of the TPBVP in the Grassmannian manifold, Gp(V). Now, with some efforts, a control function, u(t), can be determine by relating the system (11) to a family of scalar differential equation and solving the problem latter by two-point interpolation in the Grassmannian manifold, Gp(V), that affects a possible state transfer of the LTI systems and satisfies conditions that are equivalent to the controllability of the system.

4. Results

If system (10) is in the form of

0 1 0 0

( ) 0 0 1 ( ) 0 ( ),

6 11 6 1

x t x t u t

(14)

for x(t)Rn, u(t)Rm and t = T = 1sec., subject to x0 = [1 1 -2]T, xT = [0 0 0]T. Then it is possible to define (14) as a flow in the Grassmannian manifold in the form of (11) as

x(t)=Ax(t)+Bu(t), (15) for t[a,b], with boundary conditions xa=WGO and xb=WGT . A curve x(t) be a solution curve satisfying the equation (15). For system to be controllable, it is necessary and sufficient condition that the solution curve of (15), that is x(t), should be spanned by the controllability space formed by the n-vectors of the controllability matrix,Qc, Qc=[B, AB, A2B, ………, An-1B] (16) The controllability subspace <A/B> is the space spanned by the columns of matrix B with respect to the linear transformation A, [2]. We do not assume that the pair (A, B) is controllable, of course when the pair (A,B) is not controllable, only some states are transferred to the origin. Therefore solution curve of (15), x(t) = Qc(t)WG. (17) with initial and final values, from the boundary conditions, for the curve, x(t), in the Grassmannian manifold, xa =WG0 = [15 7 1]T and xb= WGT = [0 0 0]T. Now it is possible to determine a control function, u(t), by relating the system (15) to a family of scalar differential equation and solving the problem latter by two-point interpolation in the Grassmannian manifold, Gp(V), as Dnx(t) + a1D

n-1x(t) +..+ an-1Dx(t) + anx(t) = u(t) (18)

.

.

.

.

.

.

.


-1200

-1000

-800

-600

-400

-200

0

200

400

600

800

0 0.2 0.4 0.6 0.8 1 1.2

time (Sec.)

u(t

)

-35

-30

-25

-20

-15

-10

-5

0

5

10

0 0.2 0.4 0.6 0.8 1 1.2

time (Sec.)

x2(

t)

0

2

4

6

8

10

12

14

16

18

0 0.2 0.4 0.6 0.8 1 1.2

time (Sec.)

x 1(t

)

where, a1, a2, …, an are constants and operator ‘D’ represent differentiation with respect to time t. The system states can be described by equivalence between (15) and (18) as, x(t) = x1(t), Dx(t) = x2(t) ,..., D

n-1x(t) = xn(t). (19) Here, instead of solving (18) for the response x(t), we look it as a formula for a control function u(t) in terms of the response x(t) and look upon the desired state transfer of the LTI system as providing two point boundary conditions on an n-times differentiable function x(t), in the Grassmannian manifold, Gp(V). Thus by an interpolation technique a control function u(t) can be synthesize as, u(t) = -988 - 133t +7267.5t2 – 2111t3 – 4437.5t4 -669t5 (20) and the system states that satisfies the conditions that are equivalent to the controllability of the LTI system (15), in the Grassmannian manifold, Gp(V), x1(t) = 15 + 7t + 0.5t2 – 193.5t3 + 282.5t4 – 111.5t5 (21) x2(t) = 7 + t – 580.5t2 + 1130t3 – 557.5t4 (22) x3(t) = 1 – 1161t + 3390t2 – 2230t3. (23) The transfer characteristics of a control function, u(t) and the system states x1(t), x2(t) and x3(t) are shown in Fig. 3, 4, 5 and 6 respectively. 5. Conclusion

Main objective here is that the control engineers should not restrict himself to any one of the tool but should be familiar with as many as possible for analysis purpose. The interesting relationship between the differential geometry of certain curves and the controllability of LTI control systems have developed here without considering any matrix riccati equations. We have developed some interesting relationship between the curve obtained by the evolution of the continuous transformation of the initial condition under the action of the group in the Grassmannian manifold with the solution curve of TPBVP, that satisfies conditions that are equivalent to controllability of the LTI control systems. This new idea of analysis of control systems using differential geometrical approach may help us greatly in near future. Graphical results shows possible state transfer of the linear time-invariant control system in the Grassmannian manifolds, by a control force, u(t), that satisfies conditions that are equivalent to the controllability of the system. This method has the flexibility of choosing the time interval t = T sec. during which the transfer of the states from initial to final values are desired. We can also extend the same idea for analysis of linear time-varying control system.

Fig.3 Transfer char. of a control function u(t)

Fig. 4 Transfer char. of the state x1(t)

Fig. 5 Transfer char. of the state x2(t).


-150

-100

-50

0

50

100

150

0 0.2 0.4 0.6 0.8 1 1.2

time (Sec.)

x3(

t)

Fig. 6 Transfer char. of the state x3(t).

References

[1] R.E. Kalman, Y.C. Ho and K. S. Narendra, “ Controllability of Linear Dynamical Systems,” in contribution to Differential Equation. Vol. 1, pp 189-213 , John Wiley and Sons Inc., New York, 1963.

[2] B.K.Lande, “Some General problems in linear system theory”, Ph.D.Thesis submitted to Indian Institute of Technology, Bombay, India, 1985.

[3] S.D.Agashe and B.K.Lande, “A New Approach to the state transfer problem”, J.Franklin Inst., Vol.333 B, No.1, pp 15-21, 1996.

[4] Mrs.Nirmala Prakash, “Differential Geometry an integrated approach,” Tata Mc-Graw Hill Publishing Company, 1981.

[5] S. Balakumar and C. Martin, “Two point boundary value problems and the matrix Riccati Equations”, operator methods for optimal control problems, Ed. Sung J. Lee, Vol. 108, Lecture notes in Pure and Applied Mathematics, Marcell Dekkar, pp 1-37, 1987.

[6] B.F.Doolin and C.F. Martin, “Global Differential Geometry: An introduction for control engineers, NASA reference publication 1091, 1990.

[7] L.D. Drager, R.L. Foote, C. F. Martin and J.Wolper, “Controllability of linear systems, differential geometric curves in grassmannians and generalized grassmannians and riccati equation.”, An International Survey Journal on Applying Mathematics and Mathematical Applications, Vol. 16, No. 3, pp 281 – 317, Sept. 1989. (Springer link – Nov. 2004).

S.M.Deshmukh received his M.E.(Adv.Elect.) degree from the SGB Amravati University. He is currently a researcher and professor in Elect. and Tele-comm. Deptt., Prof. Ram Meghe Instt. of Tech. and Research, Badnera, Amravati. His current research interest is in the field of control systems and communication engineering. He is member of many professional bodies like IETE, ISTE, IE etc. Mrs Seema S.Deshmukh received her B.Sc.(physics) degree from Pune University in 1990and M.Sc.(physics) from North Maharastra University, Jalgaon, in 1993. She is currently a researcher and selection grade lecturer in First Year Engg. Deptt., Prof. Ram Meghe Instt. of Tech. and Research, Badnera, Amravati. Her current research interest is in the field of Mathematical Physics, Glasses etc. She is member of many professional bodies like ISTE,IE etc. R.D.Kanphade received his Ph.D. degree from SGB Amravati University in 2008.He is currently working as a Principal,Dhole-Patil Engg. College, Pune,Maharastra. His main research interest is in the field of comm..engg., control system engg. and VLSI Design. He has published number of papers in national, International journals and conferences. He is member of professional bodies like IEEE, ISTE, IETE etc. N. A. Patil received his M.Sc. (Maths) degree from SGB Amravati university, Amravati in 1986, M. Phil. from North Maharashtra University, Jalgaon in 1997 and Ph. D. degee from North Maharashtra University, Jalgaon in 2005. He is currently a researcher and selection grade lecturer in Mathematics Department, Sant Gajanan Maharaj Engineering College, Shegaon. His current research interest is in the field of Integral, Fourier, Fast Fourier, Wavelet, Laplace, Mellin Transform and Application of Digital Signal Processing. He has published number of papers in National, International Conferences and Journals. He has also published three books.


41

Strategies to Achieve Labor Flexibility in the Garment Industry

Parisima Nassirnia1, Masine Md. Tap2

1 Department of Industrial Engineering, Faculty of Mechanical Engineering, Universiti Teknologi Malaysia Skudai, Johor 81310, Malaysia

2 Department of Industrial Engineering, Faculty of Mechanical Engineering, Universiti Teknologi Malaysia Skudai, Johor 81310, Malaysia

Abstract

With the aston ishing technolog ical progr ess in industr y dur ing the current epoch, companies are facing a dynamic unpredictable market env ironment tha t m akes s urvival to a c ertain extent difficult for th em. Consequently, flexibility seems to be the bes t approach to enable companies to hav e an eff ective response to environmental threats such as economic crisis, growing competitors and m arket dem and fluctu ation. Among differen t types of fl exibility, labor flexibility seems to have a determining effect in m aking a s ystem flex ible, s ince labors are on e of the most important resources of a co mpany. T his proj ect t ries t o present a new m ethod to achieve labor flexibility as an effective strategy for garment manufacturing companies that faces market demand f luctuations. The f inal result of the pr oject shows the proposed algorithms are capable of finding the optimum strategic solution for applying labor f lexibility on th e pr oduction l ine of the case study. Keywords: Labor Flexibility, Optimization Algorithms, Simulation.

1. Introduction

The unp redictable co ndition of tod ay’s m arket environment compels manufacturing companies to switch to flex ibility to en able th em to effectiv ely respo nse to market cha nges [1]. For m aking a firm perfectly adapte d and flexi ble according t o the m arket de mand, all the elements of the syste m such as product, process, volume, material handl ing, m achine a nd l abor need to be flexible [2]. However among the many elements of flexible factors, labor flex ibility is th e most challenging facto r because of the different a spects of human behaviors t hat m ake i t as the most valuable resource and the m ost flexible elem ent of a system . Fo r th is reason more atten tion is n eeded for applying lab or flex ibility to companies in comparison to other elements of a system. In every manufacturing system there are m any p ossibilities fo r th e number of labo rs to perform work and many combinations for the skill level of labor. T hus optimization t ools are nee ded i n fi nding t he

best so lution with regard to th e labo r co st an d the productivity of th e system [1 ]. Neverth eless few researches have been done t o p ropose t he methods f or applying labo r flex ibility effectiv ely. Th is project tries t o propose so me algo rithms fo r using simulation as a planning t ool in garment i ndustry t o e valuate different labor flexibility strategies on a case stu dy and finding the optimal labor flexibility solution.

2. Literature Review

Nowadays flexibility is applied as a major competitive tool by organization t o a dapt t o the unstable a nd c hangeable environment of t he m arket [3]. The literatu re rev iew on flexibility h as b een done to present a comprehensive definition for the con cept of flex ibility an d d evelop its framework [4]. The most common types of manufacturing flexibility th at h ave been discussed in th e literatu res are machine, product, operation, material h andling, process, routing and lab or flex ibility [5],[6]. H owever labor flexibility is o ne of th e foundational factors for flex ibility pyramid of a system and consequently needs more effort to achieve [7] . Following a rev iew on th e lab or flexibility literatures, it h as been found t hat m ost o f t he labor flexibility definitions p oint to fu nctional flex ibility. For instance Ko ste (1 999) defin ed t he lab or flexibility as th e number and variety of operations that a l abor can perform with th e m inimum p enalty to th e p erformance and operation process. However labo r flex ibility can b e applied on t he firm ’s wo rkforce i n both quantitative an d qualitative app roaches. Qualitativ e ap proach of lab or flexibility o r functional labo r flexibility is th e labo r’s ability to handle different tasks and move between jobs or duties as demand changes. While quantitative approach of labor flexibility or numerical labor flexibility is the ability of c ompany t o c hange t he amount of w ork as dem and changes by changing the number of labors or the number of working hours [8], [9]. Hence, based on the numerical and fun ctional asp ect of labor flex ibility, th e m ost


42

comprehensive definition for labor flexibility is the ability of a firm to change the amount of work by changing the number of l abor or number o f w orking hours an d t he variety o f work performed b y l abor as demand c hanges with the minimum penalty to the operation processes. Permanent worke rs wit h changea ble working hours or temporary worke rs are exa mples of numerical labor flexibility. However these approaches may not always be effective a nd p ossible, si nce having m any permanent workers ar e co stly and ap plying tem porary labo rs m ay affect work qu ality and jo b standards o f the company. Thus an y sug gestion fo r ach ieving l abor flexibility strategies on a case study needs to be evaluated to find the optimized so lution with m inimum co st and withou t an y penalty to the firm’s targets and production. The objective of th is project is to in vestigate th e effect o f app lying numerical an d fun ctional lab or flex ibility o n the production t ime an d l abor co st o f a case study ( garment company) and to find the optimum solution.

3. Problem statement

Production lines of m ost g arment in dustries h ave sim ilar characteristics. They a re mostly consisted of an assembly line th at con sists o f a set o f sequ ential work stations, typically conn ected by a c ontinuous m aterial han dling system. It i s desi gned t o a ssemble com ponent pa rts o f a cloth and to perform any related operations to produce the finished cloth. There are many types for assembly lines and each type has special charac teristics. One of th e ch aracteristic o f an assembly lin e is th e d egree of au tomation [12 ]. Th e assembly lin e in m ost g arment in dustries are sem i- automated o r fully m anually. For th is reason t he ro le of labor i n pe rforming j obs i n m ost garm ent i ndustries are higher th an machines. C onsequently lab or utilization, equalize workload percentage of each operator in assembly line and labor cost a re c hallenging m anagerial i ssues i n garment i ndustry. Thi s project t ries t o pr opose som e strategies for ga rment i ndustries re garding t o t hese challenges. The first strateg y is n umerical labor flexibility that tries to d etermine the number of operators in a sim ple assembly line o f a garment in dustry an d distribute workload between th em with m inimum labor co st and minimum d elay in to tal req uired tim e for fi nishing products. The case study is th e sew ing department o f a g arment company in Malaysia, where all th e o perations for manufacturing cl oths a re se mi-automated or c ompletely manually. The p roduction qu antity and planning of product a re based o n t he cust omer’s order. Normal working hours o f t he c ompany i s 8 hours pe r day, 6

working days a wee k. C ompany has s ome perm anent workers and some temporary labors. In some cases, ba sed on the production quantity and the due time that customers request, th e working hours an d number of tem porary labors a re varied acc ordingly. For instance, ov ertime, subcontract, out so urcing a nd increasing the num ber of workers are the strategies tha t are done by company when normal work ing hours is not enou gh to fulfill d emand in the desired due tim e [10]. Th is paper uses the case study company to a ssess the effec t of applying numerical and functional algorithms to achieve labor flexibility. In this paper we assume the company receives an order for producing 3000 pieces of military clo ths within 30 working days. Acc ording t o the com panies’ rules eac h working day of co mpany is 8 hou rs as normal w orking time and maximum allowed overtime work is 2 h ours per day f or eac h worker. T herefore, t he m aximum possi ble production wo rking hou rs fo r each lab or i n sewing department during 30 days is 300 hours (30 working days × (8 + 2) hrs/day). F urthermore i t i s assum ed t hat t he labors’ sk ills are classified in to 3 levels b ased on their speed i n performing a dut y. This cl assifications a re f ull skill, semi-skill and low-skill workers that each group is able to co mplete assign ed jobs within th e stand ard tim e, 120% of st andard t ime, an d 1 40% o f s tandard t ime respectively. In ad dition, in term o f th e availability o f labors for performing operation process, the basic model is defined as a model in which for each operation one labor is always av ailable an d th is model represen ts m inimum production t ime t hat i s possi ble based on t he cu rrent production l ine of t he sewing department. The number of operations a re re quired t o produce m ilitary cl oths i n sewing de partment are 3 6 operation p rocesses, an d t he maximum required number of l abors for performing this order are 36 persons.

The aim h ere is to ap ply l abor flexibility to have t he minimum cost without any penalty to production time and customer due date. To determine the relation between cost and t ime i n thi s t ype o f c ontract p roject, we di vide t he production cost to two terms: time related cost (TRC) and time unrel ated c ost (TUC) (refer Eq uation 1). Ti me unrelated c osts are som e costs such as t he cost of raw material, fact ory re nt, depreciation, t axes which ca n be determined before starting the project and can be assumed as predictable constant cost. While time related cost is cost related to production and labor which varies according t o the to tal prod uction tim e. Hen ce for minimizing the production co st, we need to min imize th e TRC as it is shown in Equation 2. Fo r this r eason we do no t consider TUC in this research. Instead we will try t o minimize the TRC th at is a fun ction of production time an d labors’ wages (Eq.3).

1


43

TRC 2 , 3

4. Proposed labor flexibility strategies

After the model of assembly line in sewing department are built and verified using Witness simulation software, two labor flexibility strategies are applied to the model.

4.1 Numerical labor flexibility strategy

The proposed alg orithm fo r nu merical l abor flex ibility strategy is sho wn in Figure 1. Using th e pro posed algorithm (Figure 1) the experiment was conducted on the basic m odel t o i mprove i t by re ducing t he n umber o f labors.

Fig. 1 The algorithm of numerical labor flexibility strategy.

According to th is algorithm in each m odel two most id le labors will be selected and then the d uty of o ne o f them

will be added to another labor and the first labor that now does not have any job will be omitted from the model. The next st ep i s ru nning t he m odel t o t est t he resul t of l abor reduction on th e to tal production tim e. If th e to tal production t ime chan ged fav orably, t he new m odel replaces t he model and the reduction of labors will be continued in the same way until no reduction in production is possible. By perf orming t he e xperiment based o n the pr oposed algorithm, the numbers of labors were decreased from 36 to 20 l abors without any penalty to production t ime. The production time at each stage of the experiment is s hown in Figure 2 based on the number of labors in each model. As it has been shown in Figure 2 by reducing the number of labors from 36 to 20 persons the production time do not change, but after reducing t he n umber o f labor fr om 20 persons t o 19 pe rsons, t he pr oduction t ime change s dramatically. Thu s at th is po int th e ex periment was stopped and no more labor reduction is done. Figure 3 sh ows t he i nverse rel ation between number of workers a nd t he cost of workers when t he num ber of labors reduces from 36 to 20 persons. On the other hand it can be seen i n Figure 3 after re ducing t he number of labors from 20 to 19 workers, even though the number of labors re duces, labor tim e and cost i ncreases. T his is because of inc rease in the overtime wage of the labors as production time increases beyond the normal work time..

Fig. 2 Total production time versus the number of labors.

4.2 Functional labor flexibility strategy

In th is strategy th e effect of ch anging th e sk ill lev el of labors will be investigated on the op timum so lution from the numerical labor flexibility strategy with 20 full skilled labors or model No.17 and then the number of full skilled labors i n t he model will b e redu ced by su bstitution wit h low-skill or se mi-skill lab ors without any p enalty to


44

production time. The algorithm of this strategy is shown in Figure 4. In this algorithm when the substitution of labor makes significant change in production time, the skill level of worker will be increased one step higher to decrease the production time ag ain. Th is will con tinue until t he all possible skill level of all the workers were tested.

Fig. 3 Total labor cost versus the number of labors.

Fig. 4. The algorithm of functional labor flexibility strategy.

At th e end of this exp eriment, 16 co mbinations of labo rs with different skill levels were found without any penalty to the production time (refer Figure 5). For finding the best combination of labors we consider the combination with minimum labor cost (refer Figure 6). In Figure 6 the changes of labor cost in the model is shown versus the changing in t he skill level of worker. It shows that the labor cost is not constant among those models with minimum p roduction ti me an d varies as the co mbination of l abors c hanges i n t he m odels. However am ong al l o f these combinations, there is on e combination consisted of 11 low skill labor, 4 semi skill labors and 5 full skill labor which resu lted with th e min imum co st. Th erefore t his model is th e optimum co mbination of lab ors i n term of number of labor as well as the skill level of labors [12]. 5. Discussion Here we fo cus on t he statistical resu lt of the sim ulation model du ring t he di fferent st eps of ex periment an d investigate the changes in the model specifications through each step of e xperiment. The de tails of some steps of first experiment (Nu merical lab or flex ibility exp eriment) are shown in Table 1. Model No.1 shows the specification of the basic model of current sewing produ ction line. As it was mentioned before, in th e b asic m odel it is assu med th at fo r ev ery single process, one labo r is allo cated and th is lab or is always availa ble to perform process wh en it is requ ired. The labors are sorted from the busiest to the idlest labor. It can be found from Table 1 that in the basic model (Model No.1) more than hal f o f the total numbers of 36 workers have more t han 50% idle t ime. Thus t he experiment was performed by reducing the nu mber of labor s to distribute duties between labors more effectively. Comparing t he fin al statistical resu lt of lab ors i n m odel No.17 (optimized model) with others in Table 1 shows the busy time of labors that perform the same duty through all the steps of experiment do not change until the last model. For example the busy time percentage of labor011 that had been assigned to one operation from first step is the same p (71%) in all the steps of experiment. On the other hand the percentage o f busy t ime of some l abors t hat d uring t he experiment more processes we re ad ded t o t hem wer e increased. For instance t he bu sy ti me p ercentage of labor004 at fi rst is 10.5% however after assign ing two other op eration processes t o th is lab or, its b usy tim e percentage was increased to 50.33% in the final model.


45

Fig. 5. Combination of labors with different skills versus production time.

Fig. 6: Combination of labors with different skills versus labor cost.


46

Table 1: The changes in the busy time percentages of each worker on some selected models in experiment one.

Model No. 1 Model No. 7 Model No. 13 Model No. 17 No. of allocated processes

in last model.

36 Labors 30 Labors 24 labors 20 labors

Name % Busy Name % Busy Name % Busy Name

% Busy

L001 99.88 L001 99.88 L001 99.88 L001 99.88 1 L016 95 L016 95.03 L016 95.04 L026 99.32 2 L019 77.26 L019 77.2 L019 77.14 L002 95.06 2 L013 75.47 L013 75.47 L013 75.47 L016 95.05 1 L006 74.43 L006 74.43 L006 74.43 L035 83.68 2 L018 73.19 L018 73.19 L018 73.25 L019 77.09 1 L011 71 L011 71 L011 71 L013 75.47 1 L012 68.17 L012 68.17 L012 68.17 L006 74.43 1 L030 66.41 L030 66.42 L030 66.41 L018 73.25 1 L010 61.79 L010 61.78 L010 61.79 L011 71 1 L015 59.79 L015 59.78 L015 59.8 L036 70.81 5 L021 54.24 L021 54.31 L032 57.05 L012 68.17 1 L023 53.46 L023 53.45 L021 54.36 L030 66.38 1 L005 50.33 L005 50.33 L023 53.48 L010 61.82 1 L031 49.75 L031 49.76 L004 50.38 L015 59.76 1 L026 49.57 L026 49.57 L005 50.33 L032 57.04 8 L002 48.35 L002 48.35 L031 49.74 L021 54.33 1 L029 46.76 L029 46.77 L026 49.58 L023 53.5 1 L035 43.82 L035 43.82 L002 48.35 L004 50.41 3 L014 39.86 L014 39.86 L029 46.72 L005 50.33 1 L017 39.45 L017 39.49 L035 43.82 L008 26.54 L008 26.54 L014 39.86 L028 14.29 L007 19.2 L017 39.42 L003 13.87 L036 17.04 L036 31.33 L009 13.29 L028 14.29 L020 10.5 L003 13.87 L004 10.5 L032 13.48 L022 9.49 L009 13.29 L025 8.34 L020 10.5 L034 7.15 L004 10.5 L007 4.75 L036 4.36 L024 4.36 L032 3.97 L033 2.38

A comparison of the percentages of busy time of labors in the optimized model (Model No.17) with other models in the first experiment shows t hat in all m odels at least t here is o ne lab or with less th an 5 0% id le time, wh ile in the optimized m odel (model No.17) no labo r has less t han 50% idle tim e. In fact i n each step of the experiment the number of labors with more than 50% idle time decreases

smoothly until it reach es 0% in th e optimized m odel of first experiment (refer Figure 7). In this section, the result of the second experiment will be discussed. A s i t was m entioned before in t he second algorithm th e co mbinations of lab ors with d ifferent sk ill levels were ap plied to th e m odel with th e optimized


47

number of labo r (fin al resu lt of t he first strateg y, th at is, model No .17) and finally th e o ptimized model with best combination of lab ors and d ifferent sk ill lev els were found. Th e idle ti me p ercentages of labors i n t he final optimized model are presented in Table 2. By comparing the busy time p ercentages of labo rs with o ther models in both ex periments, i t can be found t hat t ask al location o f labors and work distribution between labors in this model has improved.

Fig. 7. Labors percentages in the models with more than 50% idle time

Finally th e an alytical resu lts of bo th exp eriments are summarized as follows: 1. Jo b co mbination of tho se lab ors th at their b usy ti me percentage is more th an 50% will resu lt in a si gnificant increase in total production time. Henc e for reducing the number of labors without any delay in the total production time, it is su ggested to co mbine th e jobs of labors with busy time less than 50% only. 2. Using high skilled workers who can perform jobs faster is no t always the best so lution in m inimizing produ ction time in an assem bly line. So metimes b y allo cating l ow skilled workers to th e production assembly lin e, t he total labor cost can b e d ecreased withou t any p enalty to production t ime. For finding the cheapest combination of labors with different skill levels in a p roduction assembly line, th e cycle ti me o f tho se lab ors that h ave m ore slack time (idle time) ca n be inc reased and t he effect of this increase on the to tal production ti me can be inv estigated by using simu lation. So me co mbinations do not have significant effect o n the total production time, wh ile some of them increase total production time. However by using simulation, th e op timized com bination of lab ors with the

lower co st an d l owest to tal produ ction tim e can be attained. Table 2. Specifications of labors in the optimized model (model No.17).

6. Conclusion

Two al gorithms for numerical and functional l abor flexibility were proposed. The effectiv eness of t hese algorithms was inve stigated on a case st udy (a garment company) by using simulation. The purpose of t his study is find ing th e optimized so lution for th e case stud y after testing th e resu lts of each altern ative lab or flexibility strategies o n t he p roduction time and pr oduction c ost of the case study . T he following conclusions can be d rawn from the research: 1. B y a pplying t he first st rategy ( numerical l abor flexibility) on the case st udy, the numbers of labors were reduced from 36 persons to 20 persons without any penalty to production time.


48

2. Tak ing th e o ptimal so lution after th e application of numerical flexib ility it is th en sub jected t o th e algo rithm for th e fun ctional labo r flexibility. Th e so lution was further improved when the number of full-skill labors was decreased fr om 20 to 5 pe rsons wi thout any penalty to production time. At the final stage the optimal model with the lowest labo r co st was obtained b y using 20 labors including 5 full-sk ill lab ors, 4 sem i-skill labors and 11 low-skill labo rs witho ut an y p enalty to production due date. These results indicate t hat it is possible to ac hieve optimum lab or flexibility strateg y in a p roduction assembly line of a garment factory through both numerical and fun ctional flex ibility alg orithms p roposed in th is paper. References [1] Gert. Z. U, and Sven. R , and Thorsten. V. (2004). Simulation

Approach For P lanning And Re-Assigning Of P ersonnel In Manufacturing, International Journal of Production Economics. Vol.90, pp. 256-277.

[2] Chang, A. Y.( 2004). On the measurement of labor flexibility , In ternational Engineering Management Con ference. NJ 08855-1331, United States.

[3] Vokurka, R. J., and O'Leary-Kelly, S. W. (2000). A review of empirical res earch on m anufacturing flex ibility. Journal of Operations Management. 18(4), 485-501.

[4] Oke, A. (2005). A framework for analy zing manufacturin g flexibility. International Journal of Operations & Production Management. Vol. 25 No. 10, pp. 973-996.

[5] Narain,R . Yadav, and R. C. Sarkis, and J. Cordeiro , J.J. (2000).The Strateg ic Implica tions of Flexibility in Manufacturing S ystems. International journal of Agil e Manufacturing Systems. 2/3 202-213.

[6] Koste, L. L., and Malhotra, M . K . (1999). A Theoretical Framework for Analy zing The Dimensions Of Manufacturing Flexib ility. Journal o f Operations Management.18(1), 75-93.

[7] Karuppan, C.M.(2004). Strategies to Foster Labor Flexibility. International Journal of Pro ductivity and Performance Management. Vol. 53 No. 6, pp. 532-547.

[8] Looise, J.C. & Maar ten van Riemsdijk & Frans de Lange.(1998). C ompany Labor Flexibility Str ategies In Th e Netherlands: A n Institut ional Perspective. Journals of Employee Relations. 20 ( 5), 461-482.

[9] Riley , M. and Lockwoo d, A. (1997). Strateg ies And Measurement For Workforce Flexibil ity: An Ap plication Of Functional Flex ibility In A Se rvice Se tting. I nternational Journal of Operations & Production Manag ement. 17(4 ), 413-19.

[10] Loqman, M. N. (2007) . Productivity Improvement In a Garment Manufacturing Company using Line ba lancing and Simulation., ( B.S thesis), Department of Industrial Engineering, Universit Teknologi Malaysia.

[11] Motandang, M. Z. Jambak, M. I. (2010). Soft computing in optimizing assembly lin es bal ancing. Journal of Computer Science 6 (2): 141-162.

[12] Nassirnia, P. (2009). Evalua tion of str ategies to achiev e labor flex ibility., (Master thes is), Dep artment o f Industrial Engineering, Universit Teknologi Malaysia.

Parisima Nassirnia. She received her bachelor's degree in Mechanical Engineering at Shahrood University of Technology in 2005. After 3 years of experience as a technical consultant in Iran, she joined the Faculty of Mechanical Engineering in Universiti Teknologi Malaysia (UTM) and received her Master’s degree in Industrial Engineering in 2010 from UTM. Currently she is PhD candidate and her research interests are application of modeling and simulation techniques in optimization of Manufacturing systems, Operation management, Supply chain management and logistics. Masine Md. Tap. She was born in Malaysia. She received her Bachelor’s Degree in Mechanical Engineering from Universiti Teknologi Malaysia in 1986, MPhil in Computer Aided Engineering from Herriot-Watt University, United Kingdom in 1989 and PhD. from Dundee University, United Kingdom in 1999. She is now an associate professor in the Faculty of Mechanical Engineering, Universiti Teknologi Malaysia.

IJCSI CALL FOR PAPERS JANUARY 2011 ISSUE

V o l u m e 8 , I s s u e 1

The topics suggested by this issue can be discussed in term of concepts, surveys, state of the art, research, standards, implementations, running experiments, applications, and industrial case studies. Authors are invited to submit complete unpublished papers, which are not under review in any other conference or journal in the following, but not limited to, topic areas. See authors guide for manuscript preparation and submission guidelines. Accepted papers will be published online and indexed by Google Scholar, Cornell’s University Library, DBLP, ScientificCommons, CiteSeerX, Bielefeld Academic Search Engine (BASE), SCIRUS, EBSCO, ProQuest and more. Deadline: 05th December 2010 Notification: 10th January 2011 Revision: 20th January 2011 Online Publication: 31st January 2011 Evolutionary computation Industrial systems Evolutionary computation Autonomic and autonomous systems Bio-technologies Knowledge data systems Mobile and distance education Intelligent techniques, logics, and

systems Knowledge processing Information technologies Internet and web technologies Digital information processing Cognitive science and knowledge

agent-based systems Mobility and multimedia systems Systems performance Networking and telecommunications

Software development and deployment

Knowledge virtualization Systems and networks on the chip Context-aware systems Networking technologies Security in network, systems, and

applications Knowledge for global defense Information Systems [IS] IPv6 Today - Technology and

deployment Modeling Optimization Complexity Natural Language Processing Speech Synthesis Data Mining

For more topics, please see http://www.ijcsi.org/call-for-papers.php All submitted papers will be judged based on their quality by the technical committee and reviewers. Papers that describe on-going research and experimentation are encouraged. All paper submissions will be handled electronically and detailed instructions on submission procedure are available on IJCSI website (www.IJCSI.org). For more information, please visit the journal website (www.IJCSI.org)

© IJCSI PUBLICATION 2010

www.IJCSI.org

IJCSIIJCSI

© IJCSI PUBLICATION www.IJCSI.org

The International Journal of Computer Science Issues (IJCSI) is a well‐established and notable venue

for publishing high quality research papers as recognized by various universities and international

professional bodies. IJCSI is a refereed open access international journal for publishing scientific

papers in all areas of computer science research. The purpose of establishing IJCSI is to provide

assistance in the development of science, fast operative publication and storage of materials and

results of scientific researches and representation of the scientific conception of the society.

It also provides a venue for researchers, students and professionals to submit ongoing research and

developments in these areas. Authors are encouraged to contribute to the journal by submitting

articles that illustrate new research results, projects, surveying works and industrial experiences that

describe significant advances in field of computer science.

Indexing of IJCSI 1. Google Scholar 2. Bielefeld Academic Search Engine (BASE) 3. CiteSeerX 4. SCIRUS 5. Docstoc 6. Scribd 7. Cornell's University Library 8. SciRate 9. ScientificCommons 10. DBLP 11. EBSCO 12. ProQuest

Date post:	18-Nov-2023
Category:	Documents
Upload:	independent
View:	0 times
Download:	0 times

Strategies to Achieve Labor Flexibility in the Garment Industry

Documents