Introduction to ClearPass Device Insight
Viswesh Ananthakrishnan, Senior Director, Product Mgmt.
2019 September
2@ArubaAPAC | #ATM19APAC
ARUBA SECURITY USER/DEVICE LIFE CYCLE
Aruba 360 Security
Authorize
Respond
Enforce
Act on attacks with a wide range of policy actions
Detect attacks
with range of analytics
IT access based on user/device attributes
Visibility through profiling and classification
ClearPass
Device
Insight
MonitorDiscover
CURRENT CHALLENGES IN DEVICE VISIBILITY
IT/Security teams lack visibility into devices on the network i.e. factory
controllers, medical equipment
Current toolset fails to adequately address visibility and IoT use cases
Volume, variety and the innovation of “things” means manual approaches cannot keep pace
Without comprehensive visibility, effective security and compliance is not possible
4@ArubaAPAC | #ATM19APAC
HALF OF ENTERPRISE STRUGGLING TO SECURE IOT
Source: Ponemon Institute
5@ArubaAPAC | #ATM19APAC
CLEARPASS DEVICE INSIGHT OVERVIEW
Reduces Risk by Eliminating Blind Spots
through DPI-based discovery and profiling of devices
Automatically Clusters Unknown Devices and recommends
classificationusing advanced machine learning
and crowdsourcing intelligence
Ensures Secure Accessvia seamless integration with ClearPass Policy Manager
6@ArubaAPAC | #ATM19APAC
TRADITIONAL PROFILING TECHNIQUES LACK DEVICE CONTEXT
STATIC ATTRIBUTES
NMAP | SNMP | WMI
GENERIC “WINDOWS” OR “LINUX” DEVICE
7@ArubaAPAC | #ATM19APAC
CLEARPASS DEVICE INSIGHT: FROM GENERIC TO GRANULAR DEVICE VIEW
WINDOWS DEVICE
AXIS DEVICE
AXIS SECURITY CAMERA
AXIS Q35 NETWORK CAMERA
DEEP PACKET INSPECTION (DPI)
BEHAVIORAL ATTRIBUTES
APPLICATIONSWEB SITES
PORTSPROTOCOLS
CROWD-SOURCING
MACHINELEARNING
8@ArubaAPAC | #ATM19APAC
TRADITIONAL PROFILING TECHNIQUES ARE PRONE TO MISCLASSIFICATIONS
STATIC ATTRIBUTES
MAC OUI | DHCP
Device Category: APDevice Family: MotorolaDevice Name: Motorola AP
9@ArubaAPAC | #ATM19APAC
TRADITIONAL PROFILING TECHNIQUES ARE PRONE TO MISCLASSIFICATIONS
STATIC + DYNAMIC ATTRIBUTES
MAC OUI | DHCP | TCP attributes
"dst_conns": [ "10.31.22.26:3613:tcp", "alrsprd.sharp.com:3613:tcp", "SOCALARISPRD.shcsd.sharp.com:3613:tcp", "socalarisprd.shcsd.sharp.com:3613:tcp" ], "dst_hosts": [ "alrsprd.sharp.com" ], "mac_oui": [ "001723" ], "mac_vendor": [ "Summit Data Communications" ], "ports": [ "49160", "49161", "49162", etc.]
Device Category: Medical DeviceDevice Family: AlarisDevice Name: Alaris Infusion Pump
10@ArubaAPAC | #ATM19APAC
CLASSIFIES UNKNOWN DEVICES
Device Attributes
IP/MAC Address
Application Access
Communication Protocols
Communication Frequency
Deep Packet Inspection (DPI)
MACHINE LEARNING
CROWDSOURCING
11@ArubaAPAC | #ATM19APAC
Drilling Down into Device Attributes
Discovered Devices Classify known
devices with
fingerprintsClassification based on
static, flow and
behavior based
attributes
Checks for Fingerprint
Device Identified and
Labeled
ML-basedClustering & Classification
Utilizing Machine Learning for Unknown Devices
What kind of a document is this?
How many genes does an organism
need to survive? Last week at the
genome meeting here, two genome
researchers with radically different
approaches presented complementary
views of the basic genes needed for life.
One research team, using computer
analysts to compare known genomes,
concluded that today’s organisms can be
sustained with just 250 genes, and that
the earliest life forms required a mere
128 genes. The other research mapped
genes in a simple parasite and
estimated that for this organism, 800
genes are plenty to do the job – but that
anything short of 100 wouldn’t be
enough.
Topic Modelling - Latent Dirichlet Allocation
Although the numbers don’t match
precisely, those predictions “are not all
that far apart “ especially in comparison
to the 75,000 genes in the human
genome, notes Siv Anderson of Uppsala
University in Sweden, who arrived at the
800 number. But coming up with a
consensus answer may be more than just
a genetic numbers game, particularly as
more an more genomes are completely
mapped and sequenced. “It may be a
way of organizing any newly sequenced
genome,” explains Arcady Mushegian, a
computational molecular biologist at the
National Center for Biotechnology
Information (NCBI) in Bethesda,
Maryland. Comparing an ..
What kind of a document is this?
How many genes does an organism
need to survive? Last week at the
genome meeting here, two genome
researchers with radically different
approaches presented complementary
views of the basic genes needed for life.
One research team, using computer
analysts to compare known genomes,
concluded that today’s organisms can be
sustained with just 250 genes, and that
the earliest life forms required a mere
128 genes. The other research mapped
genes in a simple parasite and
estimated that for this organism, 800
genes are plenty to do the job – but that
anything short of 100 wouldn’t be
enough.
Topic Modelling - Latent Dirichlet Allocation
Although the numbers don’t match
precisely, those predictions “are not all
that far apart “ especially in comparison
to the 75,000 genes in the human
genome, notes Siv Anderson of Uppsala
University in Sweden, who arrived at the
800 number. But coming up with a
consensus answer may be more than just
a genetic numbers game, particularly as
more an more genomes are completely
mapped and sequenced. “It may be a
way of organizing any newly sequenced
genome,” explains Arcady Mushegian, a
computational molecular biologist at the
National Center for Biotechnology
Information (NCBI) in Bethesda,
Maryland. Comparing an ..
gene 0.04
dna 0.02
genetic 0.01
Words
Biology
Topics
What kind of a document is this?
How many genes does an organism
need to survive? Last week at the
genome meeting here, two genome
researchers with radically different
approaches presented complementary
views of the basic genes needed for life.
One research team, using computer
analysts to compare known genomes,
concluded that today’s organisms can be
sustained with just 250 genes, and that
the earliest life forms required a mere
128 genes. The other research mapped
genes in a simple parasite and
estimated that for this organism, 800
genes are plenty to do the job – but that
anything short of 100 wouldn’t be
enough.
Topic Modelling - Latent Dirichlet Allocation
Although the numbers don’t match
precisely, those predictions “are not all
that far apart “ especially in comparison
to the 75,000 genes in the human
genome, notes Siv Anderson of Uppsala
University in Sweden, who arrived at the
800 number. But coming up with a
consensus answer may be more than just
a genetic numbers game, particularly as
more an more genomes are completely
mapped and sequenced. “It may be a
way of organizing any newly sequenced
genome,” explains Arcady Mushegian, a
computational molecular biologist at the
National Center for Biotechnology
Information (NCBI) in Bethesda,
Maryland. Comparing an ..
gene 0.04
dna 0.02
genetic 0.01
number 0.01
computer 0.01
information 0.03
Biology
Computer
Words Topics
What kind of a document is this?
How many genes does an organism
need to survive? Last week at the
genome meeting here, two genome
researchers with radically different
approaches presented complementary
views of the basic genes needed for life.
One research team, using computer
analysts to compare known genomes,
concluded that today’s organisms can be
sustained with just 250 genes, and that
the earliest life forms required a mere
128 genes. The other research mapped
genes in a simple parasite and
estimated that for this organism, 800
genes are plenty to do the job – but that
anything short of 100 wouldn’t be
enough.
Topic Modelling - Latent Dirichlet Allocation
Although the numbers don’t match
precisely, those predictions “are not all
that far apart “ especially in comparison
to the 75,000 genes in the human
genome, notes Siv Anderson of Uppsala
University in Sweden, who arrived at the
800 number. But coming up with a
consensus answer may be more than just
a genetic numbers game, particularly as
more an more genomes are completely
mapped and sequenced. “It may be a
way of organizing any newly sequenced
genome,” explains Arcady Mushegian, a
computational molecular biologist at the
National Center for Biotechnology
Information (NCBI) in Bethesda,
Maryland. Comparing an ..
gene 0.04
dna 0.02
genetic 0.01
number 0.01
computer 0.01
information 0.03
organism 0.04
survive 0.02
life 0.01
Biology
Computer
Living
Words Topics
Mostly about ComputersMostly about Biology
Classification using topic distribution within a document
Topic Modelling - Latent Dirichlet Allocation
Document 1 Document 2 Document 3 Document 4
18@ArubaAPAC | #ATM19APAC
Behavior features separate IOT iPads from BYOD iPads
120 devices
Labels:
74% Apple iPad (89)
26% Apple iOS Device (31)
Cluster characteristics:
Significant traffic:
types: amazon_aws, apple,
cloudflare, http2, ntp
destinations:
104.20.254/24
104.20.255/24
domains: apple.com
Vendor Apple
Floors 1-5 wall iPads BYOD iPads
Separate clusters emerge for IOT vs non-IOT behaviorAruba US IT - iPad
I P a d s c l a s s i f i e d b y R u l e s
Topic 50 Topic 48
Topic 108
Topic 168
Topic 150
19@ArubaAPAC | #ATM19APAC
Behavior features separate IOT iPads from BYOD iPads
120 devices
Labels:
74% Apple iPad (89)
26% Apple iOS Device (31)
Cluster characteristics:
Significant traffic:
types: amazon_aws, apple,
cloudflare, http2, ntp
destinations:
104.20.254/24
104.20.255/24
domains: apple.com
Vendor Apple
Floors 1-5 wall iPads BYOD iPads
Separate clusters emerge for IOT vs non-IOT behaviorAruba US IT - iPad
I P a d s c l a s s I f I e d b y R u l e s
Topic 50 Topic 48
Topic 108
Topic 168
Topic 150
6th floor IOT iPads
Topic 50
Topic 150
cnn_tx_dyn 22xgoogle_tx_dyn 21xcnn_rx_dyn 21xgoogle_rx_dyn 20xapple_location_tx_dyn 12xapple_location_rx_dyn 12xindexexchange_rx_dyn 8xkrux_tx_dyn 7xicloud_tx_dyn 7xhttp2_tx_dyn 7xkrux_rx_dyn 7xgoogle_tags_rx_dyn 7xoutbrain_tx_dyn 7xoutbrain_rx_dyn 7xhttp2_rx_dyn 7xoptimizely_rx_dyn 7xoptimizely_tx_dyn 7xicloud_rx_dyn 7xindexexchange_tx_dyn 7xgoogle_tags_tx_dyn 7xapple_rx_dyn 6xapple_tx_dyn 6x
184.51.85_rx_dyn 15x184.51.85_tx_dyn 14x172.217.9_tx_dyn 14x172.217.9_rx_dyn 14x23.46.224_tx_dyn 9x151.101.180_tx_dyn 7x151.101.181_rx_dyn 7x216.58.194_tx_dyn 7x23.77.84_tx_dyn 7x216.58.193_tx_dyn 7x104.214.30_rx_dyn 7x151.101.180_rx_dyn 7x23.46.224_rx_dyn 7x104.214.30_tx_dyn 7x13.249.55_rx_dyn 7x17.248.141_rx_dyn 7x13.249.55_tx_dyn 7x23.77.84_rx_dyn 7x17.248.141_tx_dyn 7x216.58.193_rx_dyn 7x17.253.3_rx_dyn 6x17.252.194_rx_dyn 6x17.167.192_tx_dyn 6x216.58.194_rx_dyn 6x17.252.194_tx_dyn 6x151.101.181_tx_dyn 6x17.253.3_tx_dyn 6x17.167.192_rx_dyn 6x
browser_family:Chrome_ua_stat 6xbrowser_family:Safari_ua_stat 3x
50,12,81_ops_stat 2x61,50,12,81,60_ops_stat 2x53,61,50_ops_stat 2x53,61,50,12,81_ops_stat 2x12,81,60_ops_stat 2x61,50,12_ops_stat 2x81,60,55_ops_stat 2x53,61,50,12,81,60,55_ops_stat 2x50,12,81,60,55_ops_stat 2x12,81,60,55,54_ops_stat 1x53,61,12,60,55_ops_stat 1x61,12,60,55,50_ops_stat 1x53,61,12_ops_stat 1x60,55,50,54,81_ops_stat 1x53,61,12,60,55,50,54_ops_stat 1x61,50,12,81,60,55,54_ops_stat 1x55,50,54_ops_stat 1x60,55,50_ops_stat 1x12,60,55_ops_stat 1x12,60,55,50,54_ops_stat 1x60,55,54_ops_stat 1x61,12,60_ops_stat 1x61,12,60,55,50,54,81_ops_stat 1x50,54,81_ops_stat 1x
46,47,119,121,249_op55_stat 1x47,119,121,249,252_op55_stat 1x33,43,44,46,47,119,121_op55_stat 1x33,43,44,46,47_op55_stat 1x3,6,15,31,33_op55_stat 1x3,6,15,31,33,43,44,46,47,119,121_op55_stat 1x1,3,6,15,31,33,43,44,46,47,119_op55_stat 1x15,31,33,43,44,46,47,119,121,249,252_op55_stat 1x6,15,31,33,43,44,46_op55_stat 1x43,44,46,47,119_op55_stat 1x1,3,6,15,31,33,43_op55_stat 1x44,46,47,119,121_op55_stat 1x3,6,15,31,33,43,44_op55_stat 1x6,15,31,33,43,44,46,47,119,121,249_op55_stat 1x3,6,15,31,33,43,44,46,47_op55_stat 1x31,33,43,44,46,47,119_op55_stat 1x33,43,44,46,47,119,121,249,252_op55_stat 1x31,33,43,44,46,47,119,121,249_op55_stat 1x15,31,33,43,44,46,47,119,121_op55_stat 1x15,31,33,43,44_op55_stat 1x6,15,31,33,43_op55_stat 1x44,46,47,119,121,249,252_op55_stat 1x1,3,6,15,31_op55_stat 1x31,33,43,44,46_op55_stat 1x1,3,6,15,31,33,43,44,46_op55_stat 1x6,15,31,33,43,44,46,47,119_op55_stat 1x43,44,46,47,119,121,249_op55_stat 1x15,31,33,43,44,46,47_op55_stat 1x
MSFT 5.0_op60_stat 15x
Hon Hai Precision Ind. Co.,Ltd._macoui_stat 50x
40% - App ID
10% - Dest IP 20% - DHCP fingerprint
10% - MAC vendor10% User_agent
10% DNS domains
Device document (mac = 9cd21eafdeef)
Clustering devices
ClusteringPer tenant hierarchical clustering using Euclidean distance on the LDA features
0.00336
0.00399
0.00494
0.00537
0.00910
---
0.00408
0.00188
0.00002
0.00900
0.00771
---
0.00287
0.00026
0.00012
0.00705
0.00446
---
0.00615
0.00233
0.00472
0.00397
0.00687
---
Feature Vectors
Clustering Enables Easy Labeling of Devices
23@ArubaAPAC | #ATM19APAC
ARCHITECTURE OVERVIEW
On-premises data collector (appliance or virtual) and cloud-
based analyzer
Through Deep Packet Inspection (DPI), device attributes are
extracted and metadata is sent to the cloud for analysis
Campus Branch
DEVICE INSIGHT
ANALYZER
CLOUD PLATFORM
Device InsightVirtual Collector
VDevice Insight
Hardware CollectorDevice Insight
Virtual Collector*
VDevice Insight
Hardware Collector*
* roadmap
24@ArubaAPAC | #ATM19APAC
CLOUD-ENABLED COMMUNITY CROWDSOURCING
Aruba receives the signature when
rules were used to label
Signature is made available for use by
all customers
Customer labels a device using clusters or rules
Signature is tested and validated
25@ArubaAPAC | #ATM19APAC
Multi-Vendor Switching
Multi-Vendor WLANs
ClearPass Policy ManagerAUTOMATED SEGMENTATION AND
ENFORCEMENT
Internet of Things (IoT)
BYOD and Corporate Owned
ClearPass Device InsightENHANCED DISCOVERY /
PROFILING
Bi-Directional Data Exchange
CLEARPASS POLICY MANAGER AUTOMATES SECURE ACCESS
40%Of the Global 500
140+ Ecosystem Partners
26@ArubaAPAC | #ATM19APAC
PORT-BASED DYNAMIC ROLE-BASED
StaticCamera port
Printer port
PoS port
Manual configuration of ACLs, VLANs, QoS
Automate configurations with context
PCI-compliant
Hard to scale for device type and quantity across multiple
sites
Dynamic
Flatten configurations at high scale based on user, device,
app
ENFORCED BY DYNAMIC SEGMENTATION
27@ArubaAPAC | #ATM19APAC
IOT IN HEALTHCARE
ClearPass Device InsightENHANCED DISCOVERY /
PROFILING
28@ArubaAPAC | #ATM19APAC
ClearPass Device InsightENHANCED DISCOVERY /
PROFILING
IOT IN RETAIL
ZEBRA
SES
HOW WE’RE DIFFERENT
CONTINUAL INNOVATION IN CONNECTIVITY, SECURITY, AND AI
COMPLETE VISIBILITY ACROSS THE ENTIRE INFRASTRUCTRE
AUTOMATED, MACHINE LEARNING-BASED, DISCOVERY AND
PROFILING
DYNAMIC ROLE-BASED ACCESS CONTROL
DEMO
31@ArubaAPAC | #ATM19APAC
Questions?
32@ArubaAPAC | #ATM19APAC
Thank You
Rate this sessionAccess this survey via the mobile app and let us know
what you think.
Locate this session:
- Agenda
- Select Date
- Find this session
- Click Survey
Download the Event AppGain access to the latest event information.
Scan. Play. Win.Play Now! Switch your thinking, say goodbye to the old ways and
get ready for new innovations.
Visit the Tech Playground now!
Ask Aruba - Session 1: Tuesday, 24 Sep, 2:00pm – 2:45pm- Session 2: Wednesday, 25 Sep, 1:45pm – 2:30pm - Location: Town Hall at Tech Playground
Submit your Ask Aruba questions using the mobile app now!
Join the Airheads Community Scan the QR code to sign up now!