Soft Biometric Traits for Continuous User Authentication

utterlypanoramicΑσφάλεια

30 Νοε 2013 (πριν από 3 χρόνια και 6 μήνες)

53 εμφανίσεις

Soft Biometric Traits for Continuous
User Authentication

Source


IEEE TRANSACTIONS ON
INFORMATION FORENSICS AND SECURITY,
pp.771
-
780, VOL. 5, NO. 4, DEC. 2010.


Authors


Koichiro Niinuma, Unsang Park, Member,
IEEE, and Anil K. Jain, Fellow, IEEE


Speaker

Ching
-
Hsiang Chang


Date

2011/06/08

2

Outline


Introduction


Proposed scheme


Experimental result


Conclusions

Introduction


Users are authenticated only at the initial
phase


When an user leaves the terminal, he may
forget to logout


Impostor can access the terminal in the
period


3

Introduction

4

Uses

Drinks

Impostor

User

Introduction

5

The difficulty of using biometric traits


User may turn direction of face


Frequency of observation is low


It’s annoying to keep asking user to scan
fingerprint or iris

Introduction


Keep checking biometric traits


Once verification failed, require user to
login again


6

The idea of Continuous authentication

Proposed scheme


Hard


PCA
-
based face feature


Soft biometric:


Soft face (face color)


Clothes color


7

Biometric traits used in proposed scheme

Proposed scheme


Examples


Color of clothes


Color of skin


Features:


No preregistration is required


Frequency of observation is high


8

The proposed solution: soft biometric traits

Mode transition

9

Initial login
authentication

Continuous
Authentication

Enrollment
Template Update

Re
-
login
Authentication

Formulas

10


Soft face score



Hard face score



Clothes score



Soft total score


Formulas (continue)

11


Re
-
login score


where denotes a time decaying function with deciding the
decay rate ( )

k
0
k

12

Experimental result (1/4)

13

Experimental result (2/4)

14

Experimental result (3/4)

15


• False Reject (FR): The system identifies incorrectly that a user is not in the
camera’s field of view even though the user is still in front of the camera. False
rejects lower the usability of the system.

• False Accept (FA): The system wrongly identifies an impostor as the legitimate
user. False accepts lower the security of the system.

Experimental result (4/4)

16

17

Conclusions


Simple and robust




Bhattacharyya coefficient

18


A. Bhattacharyya, “On a measure of divergence between two statistical

populations defined by their probability distributions,” Bull. Calcutta

Math. Soc., vol. 35, pp. 99

109, 1943.

1
1
1
(,)
where , is feature vector;
is length of vector
EX:
[0.2,0.3,0.5]
[0.6,0.2,0.2]
(,) 0.2 0.6 0.3 0.2 0.5 0.2
=0.95
D
i i
i
s a b a b
a b
D
a
b
s a b

 


     

Re
-
login conditions

19