KIANOOSH MOKHTARIANMOHAMMAD-SADEGH FARAJI

Hybrid Authentication

Introduction

Authentication Explicit Implicit

Previous works Biometric authentication Regarding individual elements Mostly using a modeling method Try to authenticate users by characteristics

System Architecture

Learning Algorithm

User model Authentication module

Recent user behavior

Output

User credential

Intuition for our work

A combination of characteristics is uniqueCombination of characteristics is not

forgeableShort-term behavior sometimes differs

dramatically from long-term one.Short term behavior modeling should store

more detailsLeveraging multiple user behavior modeling Short term user behavior ought to be

considered periodically not event driven

Algorithm

I: recent user behaviorIf I is similar short term user behavior and does not violate long term behavior

update bothIf I is similar to long term behavior and does not violate short term user behavior

update bothElse

ask for credential

Modeling

All features are independentEach feature is considered as a random

variableScore is calculated independently for each

featureAn ensemble classifier will decide based on

these scoreGaussian Mixture Model for GPSBayesian belief network for long term

modeling K-nearest neighbor for short term modeling

7

Short term modelingStore all input data in the training

set

For each pattern in the test set

Search for the K nearest patterns to the input pattern using a Euclidean distance measure

For classification, compute the confidence for each class as Ci /K,

(where Ci is the number of patterns among the K nearest patterns belonging to class i.)

The classification for the input pattern is the class with the highest confidence.

8

Ensemble Classifier

OriginalTraining data

....D1D2 Dt-1 Dt

D

Step 1:Create Multiple

Data Sets

C1 C2 Ct -1 Ct

Step 2:Build Multiple

Classifiers

C*Step 3:

CombineClassifiers

Evaluation

Training period Store features over a week and detemine user model Call pattern over time

Authentication phase Calculate authentication score over last few hours and

figure delta= f- fp

If delta is acceptable, update probabilty density function from last few hours

Else reject the request

User modeling experiment

User 1

Satureday Sunday Monday Tuesday Wendneseday Thurseday Friday0

1

2

3

4

5

6

7

8:00 AM9:00 AM10:00 AM11:00 AM12:00 PM

User modeling experiment

User 2

Satu

reda

y

Sund

ay

Mon

day

Tues

day

Wen

dsed

ay

Thur

seda

y

Friday

0

1

2

3

4

5

6

7

8

8:00 AM9:00 AM10:00 AM11:00 AM12:00 PM

Authentication score - Calls

8:00 AM 9:00 AM 10:00 AM

11:00 AM

12:00 PM

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Score

ScoreUser 1

User 2

Authentication score - GPS

8:00 AM 9:00 AM 10:00 AM 5:00 PM0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Score

ScoreUser 1

User 2

User behavior modeling

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 9.5 100

0.2

0.4

0.6

0.8

1

1.2