Text Input Biometric System Design for Handheld Devices
Naif Alotaibi, Emmanuel Pascal Bruno, Michael Coakley, Alexander Gazarov, Vinnie Monaco, Stephen Winard, Filip Witkowski, Alecia Copeland, Peter Nebauer, Christopher Keene, and Joshua Williams
Security of handheld devices• Handheld devices play a major role in our personal and business activities.
• Securing data on the devices is critical• Currently, front-line authentication
measures are used (ex. Passwords)
Keystroke biometric authentication• Identifying users based on typing patterns• Implicit authentication with minimal user
involvement. • The keystroke biometric system at Pace
University is an effective authentication measure
• Implementing and investigate the viability of PKBS on handheld devices
Virtual Keyboards - iOS
• 2007 – the first iPhone• Almost the same keyboard• Special characters• Autocomplete and dictionary
Android keyboards - types
• 2009 – first virtual keyboards in Android 1.5 Cupcake
• Dictionary and Autocomplete• Special characters• Swype
OS Choice: Android
Raw data capture
Typical data
The key code of the touched key
Key press time
Key release time
Touch screen data
Exact touch coordinates in pixels
Finger pressure on the screen
Size and shape of the touched area
Configuration-based data
Current keyboard layout (QWERTY or one of the symbol
layouts)
Current screen orientation
Sensor-based data
Current accelerometer
values indicating the position of the
device
The change of the accelerometer
values since the last measured value
System architecture
BioKeyboard (IME service)
Biometric event
Event buffer SQLite database
Data file
Network
Settings activity
Data CollectedData Type DescriptionAction press or release
Entity code of pressed key
Keyboard Type QWERTY, or symbol layouts
Orientation portrait or landscape
Time timestamp of the event
Coordinates of the touch position
Pressure of the touch
Touch major/minor Tool major/minor
size of the clicked key
Screen data pixel density (dots per inch) & width and height (pixels)
Sensor-based data rotation (X, Y and Z) and acceleration (X, Y and Z)
Session data session ID, session time and user name.
Experimental Results
Conclusions• implemented a software keyboard system to
capture biometric events
• System allows us to run experiments, collect data, and extract features to authenticate users.
Future Work• Developing the feature vector.
• System enhancements: – Capturing gesture/Swype input.– Track spelling suggestions.
Thank You