Yanushkevich S.N., Wang P.S.P., Gavrilova M.L., Srihari S.N. (eds.) Image Pattern Recognition. Synthesis and Analysis in Biometrics
Подождите немного. Документ загружается.
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
238 Synthesis and Analysis in Biometrics
Fig. 9.8. ROC curves for the four statistical features.
Table 9.2. The recognition rate of different
combination rules.
Fusion rules Identification rate (%)
Sum Rule (Eq. 11) 96.99
Max Rule (Eq. 12) 97.32
Min Rule (Eq. 13) 96.99
Medin Rule (Eq. 14) 97.64
Majority Voting (Eq. 15) 97.51
collected for testing. It seems that the testing set is too small to cover
the distribution of all palmprints. An average recognition rate 91% was
achieved by the technology proposed in
[
27
]
, which involved a hierarchical
palmprint recognition fashion. The global texture energy features were used
to guide the dynamic selection for a small set of similar candidates from
the database at coarse level for further processing. An interesting point
based image matching was performed on the selected similar patterns at
fine levels for the final confirmation. Nevertheless, the recognition rate of
our method is more efficient, as illustrated in Table 9.3.
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
Palmprint Identification by Fused Wavelet Characteristics 239
Table 9.3. Comparison of different palmprint recognition methods.
Method Feature Points Hierarchical Identification Our Method
Database 30 Samples 200 Samples 3056 Samples
Features Feature points Global texture feature Statistical features
Accuracy 95% 91% 97.64%
9.6. Conclusions
The chapter proposed a wavelet-based palmprint identification method.
Using the wavelet transform to decompose the normalized sub images and
extracting statistical features from the sub-bands, four features (mean,
energy, variance and kurtosis) are tested. The feature of energy provides
the best performance at all potential operating points. It can operate at
high genuine acceptance rate of 94% and low false acceptance rate of 0.05%.
Its equal error rate is 4.2%. In order to get a higher recognition accuracy,
fives kinds of decision fusion strategies are used to combine the outcomes
of each classifier, and the testing results show that the Median rule and the
Majority voting rule can greatly improve the identification performance. A
satisfied recognition rate (97.64%) can be achieved by using the Median
rule.
Acknowledgments
The work is partially supported with NSFC Project (90209020) funds from
the China Government, the UGC/CRC & CERG funds from the Hong Kong
SAR Government and the central fund from The Hong Kong Polytechnic
University.
Bibliography
1. Jain, R. Bolle and S. Pankanti, Biometrics: Personal Identification in
Networked Society (Kluwer Academic Publishers, 1999).
2. D. Zhang,Automated Biometrics: Technologies and Systems (Kluwer
Academic Publishers, 2000)
3. A.K. Jain, A. Ross and S. Prabhakar, An introduction to biometric
recognition, IEEE Transactions on Circuits and Systems for Video
Technology, Special Issue on Image- and Video- Based Biometrics, vol. 14,
no. 1, January (2004)
4. R. Sanchez-Reillo, C. Sanchez-Avilla and A. Gonzalez-Marcos, Biometric
identification through hand geometry measurements, IEEE Transactions on
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
240 Synthesis and Analysis in Biometrics
Pattern Analysis and Machine Intelligence, vol. 22, no. 18, pp.1168-1171
(2000)
5. D. Zhang, Palmprint Authentication, (Kluwer Academic Publishers, 2004)
6. Jain, H. Lin, and R. Bolle, On-Line Fingerprint Verification, IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 4
(1997)
7. W. Shi, G. Rong, Z. Bain and D. Zhang, Automatic Palmprint Verification,
International Journal of Image and Graphics, vol. 1, no. 1, pp. 135-152 (2001)
8. D. Zhang and W. Shu, Two Novel Characteristics in Palmprint Verification:
Datum Point Invariance and Line Feature Matching, Pattern Recognition,
vol. 32, no. 4, pp. 691-702 (1999)
9. W. Shu and D. Zhang, Automated Personal Identification by Palmprint,
Optical Engineering, vol. 37, no. 8, 2659-2362, (1998)
10. W. Li, D. Zhang, Palmprint Identification by Fourier Transform,
International Journal of Pattern Recognition and Artificial Intelligence,vol.
16, no. 4, pp. 417-432 (2002)
11. D. Zhang, W.K. Kong, J. You and M. Wong, On-line palmprint identification,
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25,
no. 9, pp. 1041-1050 (2003)
12. Yu Tao, Ernest C.M. Lam, Yuan Y. Tang, Feature Extraction Using Wavelet
and Fractal, Pattern Recognition, 22, pp. 271-287 (2001)
13. His-Chin Hsin, Texture Segmentation Using Modulated Wavelet Transform,
Transactions on Image Processing, 9, pp. 1299-1302 (2000)
14. A.K. Jain, S. Prabhakar, L. Hong and S. Pankanti, Filterbank-based
fingerprint matching, IEEE Transactions on Image Processing, vol. 9, no.
5, pp. 846-859 (2000)
15. Vasily Strela, Peter Niels Heller, The Application of Multiwavelet Filterbanks
to Image Processing, IEEE Transactions on Image Processing, 8, 548-563
(1999)
16. Q. Xiao, H. Raafat, Fingerprint Image Postprocessing: A Combined
Statistical and Structural Approach, Pattern Recognition, vol. 24, no. 10,
pages 985-992, (1991)
17. L. Zhang and D. Zhang, 2004, Characterization of Palmprints by Wavelet
Signatures via Directional Context Modeling, IEEE Transactions on System,
Man, and Cybernetics Part-B, Volume: 34, Issue: 3, 1335-1347 (2004)
18. Thierry Blu and Michael Unser, Wavelets, Fractals, and Radial Basis
Functions, IEEE Transactions on Signal Processing, VOL. 50, NO. 3, 543-553
(2002)
19. M. Tico, P. Kuosmanen, J. Saarinen, Wavelet Domain Features for
Fingerprint Recognition, Electronics Letters, 37, 21-23 (2001)
20. Yvette Mallet, Danny Coomans, Jerry Kautsky, and Olivier De Vel,
“Classification Using Adaptive Wavelets for Feature Extraction, IEEE
Transactions on Pattern Analysis And Machine Intelligence, VOL. 19, NO.
10 (1997)
21. Josef Kittler, Mohamad Hatef, Robert P.W. Duin, and Jiri Matas, On
Combining Classifiers, IEEE Transactions on Pattern Analysis and Machine
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
Palmprint Identification by Fused Wavelet Characteristics 241
Intelligence, vol. 20, pp. 226-239 (1998)
22. John Sheffeld, The Future of Fusion, Nuclear Instruments and Methods in
Physics Research, A 464, 33-37 (2001)
23. Ravi P. Ramachandrana, Kevin R. Farrellb, Roopashri Ramachandrana,
Richard J. Mammonec, Speaker recognition-general classifier approaches and
data fusion methods, Pattern Recognition, 35, 2801-2821 (2002)
24. Andrew Senior, A Combination Fingerprint Classifier, IEEE Transactions on
Pattern Analysis and Machine Intelligence, vol. 23, pp. 1165-1174 (2001)
25. Lei Xu, Adam Krzyzak, and Ching Y. Suen, Methods of Combining
Multiple Classifiers and Their Applications to Handwriting Recognition,
IEEE Transaction on System, Man, and Cybernetics, vol. 22, no. 3, pp.
418-435 (1992)
26. N. Duta, Anil K. Jain, K.V. Mardia, Matching of palmprint, Pattern
Recognition Letters, pp. 477-485 (2002)
27. J. You, W. Li and D. Zhang, Hierarchical Palmprint Identification via
Multiple Feature Extraction, Pattern Recognition, vol. 35, no. 4, pp. 847-
859 (2002)
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
This page intentionally left blankThis page intentionally left blank
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
Chapter 10
Behavioral Biometrics for Online Computer User
Monitoring
Ahmed Awad E. Ahmed
∗
, Issa Traor´e
†
Information Security and Object Technology Laboratory,
Department of Electrical and Computer Engineering,
University of Victoria, Canada
∗
aahmed@ece.uvic.ca
†
itraore@ece.uvic.ca
Traditional biometrics technologies such as fingerprints or iris recognition
systems require special hardware devices for biometrics data collection.
This makes them unsuitable for online computer user monitoring, which
to be effective should be non-intrusive, and carried out passively.
Behavioural biometrics based on human computer interaction devices
such as mouse and keyboards do not carry such limitation, and as such
are good candidates for online computer user monitoring. We present in
this chapter artificial intelligence based techniques that can be used to
analyze and process keystroke and mouse dynamics to achieve passive
user monitoring.
Contents
10.1. Introduction ................................... 244
10.2. BiometricsModesandMetrics......................... 244
10.3. Mousedynamics................................. 247
10.3.1. Overview................................ 247
10.3.2. DetectionProcess........................... 248
10.3.3. SilenceAnalysis............................ 251
10.4. KeystrokeDynamics .............................. 253
10.4.1. Overview................................ 253
10.4.2. Free Text Detection Using Approximation Matrix Technique . . . 254
10.4.3. Free Text Detection Based on Keyboard Layout Mapping . . . . 257
10.4.4. Free Text detection Based on Sorted Time Mapping . . . . . . . . 259
10.5. Conclusion.................................... 261
Bibliography ....................................... 262
243
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
244 Synthesis and Analysis in Biometrics
Glossary
FA R — False Acceptance Rate
FRR — False Rejection Rate
FTE — Failure to Enroll
FTC — Failure to Capture
MLP — Multi-Layer Perceptrons
10.1. Introduction
Biometrics can be defined as a set of distinctive, permanent and
universal features recognized from human physiological or behavioural
characteristics
[
17
]
,
[
18
]
. As such, biometrics systems are commonly
classified into two categories: physiological biometrics and behavioural
biometrics. Physiological biometrics, which include finger-scan, iris-scan,
retina-scan, hand-scan, and facial-scan use measurements from the human
body. Behavioural biometrics such as signature or keystroke dynamics use
measurements based on human actions
[
3
]
,
[
4
]
,
[
13
]
,
[
14
]
. Due to their
strong variability over time, so far, behavioural biometric systems have been
less successful compared to physiological ones
[
3
]
. Despite such limitation,
behavioural biometrics such as mouse dynamics and keystroke dynamics,
carry the greatest promise in the particular field of online computer user
monitoring
[
7
]
. For such application, passive or non-intrusive monitoring is
essential. Unfortunately most biometrics systems require special hardware
device for biometrics data collection, restricting their use to only networks
segments where such devices are available. Behavioural biometrics such as
mouse dynamics and keystroke dynamics are appropriate for such context
because they only require traditional human-computer interaction devices.
In this chapter, we present some techniques for extracting and analyzing
mouse and keystroke dynamics data for online computer user monitoring.
While the use of mouse dynamics for online monitoring is straightforward,
the use of keystroke dynamics for such purpose faces the important
challenges underlying the need for free-text detection.
10.2. Biometrics Modes and Metrics
Biometric systems operate in two modes, the enrollment mode and the
verification/identification mode. In the first mode, biometric data is
acquired using a user interface or a capturing device, such as a fingerprints
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
Behavioral Biometrics for Online Computer User Monitoring 245
scanner. Raw biometric data is then processed to extract the biometric
features representing the characteristics, which can be used to distinguish
between different users. This conversion process produces a processed
biometric identification sample, which is stored in a database for future
identification/verification needs. Enrolled data should be free of noise and
any other defects that can affect its comparison with other samples. In
the second mode, biometric data is captured, processed and compared
against the stored enrolled sample. According to the type of application,
a verification or identification process will be conducted on the processed
sample as follows.
Verification process: conducts one-to-one matching by comparing the
processed sample against the enrolled sample of the same user.
For example, user authentication at login: the user declares his
identity by entering his login name. He then confirms his identity
by providing a password and biometric information, such as his
signature, voice password, or fingerprint. To verify the identity,
the system will compare the user’s biometric data against his
record in the database, resulting with a match or non-match.
Identification process: matches the processed sample against a large
number of enrolled samples by conducting a 1 to N matching to
identify the user; resulting in an identified user or a non-match.
In order to evaluate the accuracy of a biometric system, the following
metrics must be computed:
False Acceptance Rate (FAR), the ratio between the number of occurrences
of accepting a non-authorized user compared to the number of access
trials.
False Rejection Rate (FRR), the ratio between the number of false alarms
caused by rejecting an authorized user compared to the number of
access trials.
Failure to Enroll (FTE), the ratio characterizing the number of times the
system is not able to enroll a user’s biometric features. This failure is
caused by poor quality samples during enrollment mode.
Failure to Capture (FTC), the ratio characterizing the number of times
the system is not able to process the captured raw biometric data and
extract features from it. This occurs when the captured data does not
contain sufficient information to be processed.
FAR and FRR values can vary significantly depending on the
sensitivity of the biometric data comparison algorithm used in the
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
246 Synthesis and Analysis in Biometrics
verification/identification mode; FTE and FTC represent the sensitivity
of the raw data processing module.
In order to tune the accuracy of the system to its optimum value, it
is important to study the effect of each factor on the other. If the system
is designed to minimize FAR to make the system more secure, FRR will
increase; on the other hand, if the system is designed to decrease FRR
by increasing the tolerance to input variations and noise, FAR increases.
By considering the current utilization of biometrics in the market, we
note that they are widely used for secure identification and verification
purposes. Other security systems, such as intrusion detection systems,
do not use this technology because the application requires continuous
monitoring and recording of biometric information, which can sometimes
be unavailable during the detection process. Mouse dynamics biometrics,
which we introduce in this paper, happens to be a good fit for such an
application
[
1
]
.
Mouse dynamics biometric system does not need any training,
operating totally transparent from the user, as its enrollment mode
does not require the user to present any particular biometric
information by conducting specific actions.
Figure 10.1 shows a generic architecture, which covers the components
involved in the implementation of a behavioural biometric based detector.
The flow of control and data is shown for the three possible scenarios:
enrolling a user, verifying his identity, and identifying an unknown user.
Fig. 10.1. Generic architecture of a behavioural biometric system.
April 2, 2007 14:42 World Scientific Review Volume - 9in x 6in Main˙WorldSc˙IPR˙SAB
Behavioral Biometrics for Online Computer User Monitoring 247
10.3. Mouse dynamics
10.3.1. Overview
In contrast with keystroke dynamics, which has been widely studied in
computer security, previous works on mouse dynamics have, so far, been
limited to user interface design improvement
[
16
]
,
[
15
]
,
[
12
]
.Inparticular,
mouse movement analysis has been the purpose of extensive research works.
Studies have been conducted to establish the applicability of Fitts’ law in
predicting the duration of a movement to a target based on the size of the
target and the distance from the starting point to the target
[
16
]
. According
to Fitts’ law, the mean movement time for a movement with distance A to
a target with width W is as follows: MT = a + b(log
2
(2A/W )) where a and
b are empirically determined parameters
[
16
]
.
Researches conducted on mouse dynamics have focused mainly on the
formalization of the measured data after fixing the environment variables.
In our research, we target the biometric identification problem by focusing
on extracting the behavioural features related to the user and using these
features in computer security. Mouse dynamics is a new behavioural
biometric introduced at the Information Security and Object Technology
(ISOT) research lab at the University of Victoria
[
1
]
,
[
2
]
. Mouse dynamics
can be described as the characteristics of the actions received from the
mouse input device for a specific user, while interacting with a specific
graphical user interface. The raw data collected for each mouse movement
consist of the distance, time,andangle. We broadly refer to the angle as
the direction of movement. Silence periods between mouse movements also
carry valuable information that can be used for biometric recognition.
The user characteristics can be described by a set of factors generated
as a result of analyzing the recorded mouse actions. Those factors represent
the Mouse Dynamics Signature, which can be used in verifying the identity
of the user.
The architecture described in Fig. 10.1 can be applied for the imple-
mentation of any detector based on this biometric. Figure 10.2 shows
sample data provided to the behavior-modeling component for processing.
Each point in the figure represents a movement of the mouse with a specific
distance, which was completed in a specific time. By examining the output
of the data collection and processing component and comparing it to its
output for different sessions for the same user, one can find a pattern
characterizing the data. In some cases one can even differentiate between
readings for two different users.