Privacy-preserving and tokenless chaotic revocable face authentication scheme

Document Sample
Privacy-preserving and tokenless chaotic revocable face authentication scheme Powered By Docstoc
					Telecommun Syst
DOI 10.1007/s11235-010-9314-2

Privacy-preserving and tokenless chaotic revocable face
authentication scheme
Muhammad Khurram Khan · Khaled Alghathbar ·
Jiashu Zhang

© Springer Science+Business Media, LLC 2010

Abstract With the large-scale proliferation of biometric        1 Introduction
systems, privacy and irrevocability issues of their data have
become a hot research issue. Recently, some researchers         Recently, Biometrics-based person authentication systems
proposed a couple of schemes to generate BioHash, e.g.          have attracted much attention in implementing the security
PalmHash, however, in this paper, we point out that the pre-    of practical applications e.g. access control, e-commerce,
vious schemes are costly in terms of computational com-         and computer login, etc. [1]. Biometrics, i.e. something you
plexity and possession of USB tokens to generate pseudo-        are, has shown its superiority over the traditional authenti-
random numbers. To overcome the problems, we propose a          cation systems of passwords [something you know] and to-
novel chaotic FaceHashing scheme which preserves privacy        kens [something you have]. Biometrics data of a person is
of a biometric user. The presented scheme does not need         permanent, e.g. fingerprint or face, and cannot be changed
users to posses USB tokens in generating pseudorandom           after a certain age. In case if biometric template or data is
sequences, which is a cost-effective solution. Besides, our     compromised then it can be misused to impersonate a valid
scheme minimizes the system complexity with simple op-          user, which could effect the integrity of the whole security
erations to attain the FaceHash. Experimental results show      system. The only remedy is to replace the theft data to an-
that the proposed scheme is efficient, secure, and revocable     other biometrics template. But a person has limited num-
in case FaceHash is theft or compromised.                       ber of biometrics traits e.g. one face and two irises, so it
                                                                is not a feasible solution. It is very important to find an al-
                                                                ternative/substitute solution for the non-revocable biometric
Keywords Face · Biometrics · Chaos · Authentication ·           technology.
FaceHash · Security · Privacy                                      To solve this problem, some researchers proposed a new
                                                                approach that was named as BioHashing [2–8] that com-
                                                                bines a tokenized random number and biometrics features.
                                                                First, Connie et al. [2, 3] combined biometric palmprint fea-
M.K. Khan ( ) · K. Alghathbar
                                                                tures with a set of pseudorandom data to generate a unique
Center of Excellence in Information Assurance (CoEIA),
King Saud University, Riyadh, Kingdom of Saudi Arabia           discretized code for every person. They performed inner
e-mail:                                     product between palmprint features and pseudo random key
                                                                and named their scheme as PalmHashing, which is the con-
K. Alghathbar
                                                                cept of palmprint + hashing technique.
Information Systems Department, College of Computer and
Information Sciences, King Saud University, Riyadh,                Teoh-Ngo [4] proposed a two-factor authenticator based
Kingdom of Saudi Arabia                                         on iterated inner products between tokenized pseudo-random
e-mail:                                  number and the user specific fingerprint feature, which gen-
                                                                erated from the integrated wavelet and Fourier-Mellin trans-
J. Zhang
Sichuan Key Lab of Signal & Information Processing,             form, and hence produce a set of user specific compact code
Southwest Jiaotong University, Chengdu, Sichuan, P.R. China     that coined as BioHashing.
                                                                                                                M.K. Khan et al.

    Teoh-Ngo [5] used the same concept to generate cance-        scheme is composed of extracting the features from the cap-
lable biometrics by human face. They proposed a face hash-       tured biometrics face images to generate the face feature set,
ing technique by performing the iterated inner products be-      Generating chaotic pseudorandom numbers without using
tween tokenized pseudorandom number and face features to         the tokens, transforming the pseudorandom numbers into or-
produce a specific code, which they named as FacceHash.           thonormal vectors, performing inner products and defining
    Recently, Maio-Nanni [6] and Nanni-Lumini [7] identi-        the threshold, and finally the verification of an identity.
fied that the above cited schemes show lower performance             If the biometric data generated by our scheme is com-
if an imposter B steals the pseudorandom numbers/key of          promised, then user can be issued a new FaceHash and the
a user A and tries to authenticate or impersonate as A. As       same user can use different FaceHashes for different array of
a consequence, Maio-Nanni [6] proposed a multimodal fu-          applications e.g. National database, healthcare, and banking.
sion of face and pseudorandom numbers, and claimed that          So our scheme eliminates the fear of using biometrics for the
their scheme can prevent the attacks in the case if B steals     cross matching applications [1], and solves the problems of
the pseudorandom numbers of A. Nanni-Lumini [7] also ex-         non-revocable biometrics [8]. Moreover, proposed scheme
tended their work on the basis of same idea by using hu-         is general in nature and can be applied on any biometrics
man signatures as a behavioral biometrics. They showed           e.g. fingerprint, iris, or signature to generate BioHash.
that fusion of online signatures by combining with the to-          Rest of the paper is organized as follows: Sect. 2 presents
kenized pseudorandom numbers, generated by Blum-Blum-            the proposed scheme, Sect. 3 elaborates the experimental re-
Shub technique, could solve the security problem of previ-       sults and discussion, and at the end, Sect. 4 concludes the
ous schemes [2–7] when an impostor steals the pseudoran-         findings of this paper.
dom numbers of a legitimate user.
    However, in this paper, we point out that the all the pre-   2 The proposed scheme
vious techniques use costly tokens e.g. USB to generate the
pseudorandom numbers, which is a cost effective solution.        The presented FaceHashing scheme is shown in Fig. 1 and
All those schemes save the seed value on token and if the        in the following subsections; we elaborate our scheme in
token is compromised or stolen, an adversary can use it as       detail.
a valid user identity. Thus, those schemes suffer from the
                                                                 2.1 Face feature vector generation
problems of tokens because they can be lost, shared with
others, and can be duplicated, etc. [17]. We also assume         The proposed scheme starts with preprocessing face images
that if these mentioned schemes are to be implemented in         to generate feature vector. A 2D face image can be viewed as
some sophisticated applications e.g. Automatic Teller Ma-        a factor in the image space. Two sample face images, from
chine (ATM) then, users cannot use their tokens to gener-        ORL database [9], used in our experimentation are shown
ate the pseudorandom numbers because of the limitations in       in Fig. 2. A face gray image X with size M × N can be
ATM machines. Besides, it is not a feasible solution to is-      represented as:
sue each user a specific USB to perform transactions on the
ATMs.                                                            X = {x(i, j ), 0 ≤ i < M, 0 ≤ j < N }
    For the extension and improvement upon the related             x(i, j ) ∈ {0, 1, . . . , 2L − 1}                        (1)
schemes, we propose an efficient and Tokenless technique
in order to strengthen the privacy of biometric data. Our        where L is the number of bits to represent a pixel.

Fig. 1 Schematic diagram of the proposed scheme
Privacy-preserving and tokenless chaotic revocable face authentication scheme

   For the feature extraction, we use a well-known tech-                 and symmetric property etc. These properties are signifi-
nique called principal component analysis (PCA), also                    cant in generating a sequence of independent and identically
known as Eigenface for face recognition [10–12]. The ma-                 distributed binary random variables. The Jacobian Elliptic
jor objective of PCA is to project the high dimensional vi-              Chebyshev Rational Maps are rational functions defined as
sual stimuli i.e. face images into a lower dimensional space.            follows [13]:
PCA is an optimal method for dimensionality reduction in                    Let p be a positive integer, w ∈ [−1, 1] be a real number,
the sense of mean-square error (MSE) [11, 12].                           and k ∈ [0, 1] be a real number called modulus, then:
   Suppose the recognition space is y = {y1 , y2 , . . . , yn }T
and feature set is x = {x1 , x2 , . . . , xm }T , then the recogni-      Rp+1 (w, k) =                                         Rp (w, k)
tion space becomes y = AT x, where AT is the transpose                                     1 − k 2 (1 − Rp (w, k)2 )(1 − w 2 )
matrix of A, which can be said transfer matrix, and n and                                  − Rp−1 (w, k)                                   (5)
m are dimensions of recognition space and feature vector,
respectively. The covariance of recognition space y can be               where: p = 0, 1, 2, . . . , R0 (w, k) = 1 , and R1 (w, k) = w.
shown as:                                                                   The random vector generated by (5) is consisted of real
            1                                                            number entries distributed in the interval [−1, 1], which we
σ 2 (A) =                  ¯        ¯
                      (y − y)T (y − y) = AT   A                  (2)     show as {β1 , β2 , β3 , . . . , βm }.
                                                                         2.3 Vector orthonormalization
        ¯                  ¯
where y is the average of y ∈ Y and is the covariance ma-
trix of learning patterns, which can be mathematically rep-
                                                                         After generating the pseudorandom vector, we transform the
resented as:
                                                                         basis {β1 , β2 , β3 , . . . , βm } into an orthonormal vector set by
      1                                                                  the Gram-Schmidt process. The proof of Gram-Schmidt or-
  =                  ¯        ¯
                (x − x)T (x − x)                                 (3)
      n                                                                  thonormalization can be shown as follows:
          x∈X                                                                       def
                                                                            Let X = {β1 , β2 , β3 , . . . , βm } be a linearly independent
         ¯                    ¯
where x is the average of x ∈ X. The above equation                      set of a n × 1 pseudorandom vectors with m ≤ n. If m < n
can be used to solve the following eigenvalue problem                    then X may be filled out to a maximally linearly independent
as: AT A = , where         is a diagonal eigenvalue matrix               vector set. Hence it is assumed that m = n, where X is a
that has eigenvalues i.e. = diag{μ1 , μ2 , . . . , μ3 } and ∈            basis for C n . By Mathematical induction, we can define:
R N ×N . And the cumulative proportion can be computed as:
          m                                                              u1 =       ,
          i=1 μi                                                                 β
F=        n        × 100                                         (4)
          i=1 μi                                                                              k

where μi is the ith element eigenvalue of the diagonal ma-               wk+1 = xk+1 −             (xk+1 , ui )ui ,                        (6)
trix . Finally, the eigenvalues in decreasing order can be                                   i=1

written as: μ1 ≥ μ2 ≥ · · · ≥ μm ≥ · · · ≥ μn and normalized                         wk+1
                                                                         uk+1 =           ,        k = 1, . . . , n − 1
face feature vector can be represented as α that is normal-                          wk+1
ized to a n × 1 vector.
2.2 Chaotic pseudorandom vector generation                                 (i) Since X is linearly independent x1 > 0 and so u1 is
                                                                               well defined. Furthermore, u1 = 1.
Sequences of independent and identically distributed (i.i.d.)             (ii) Mathematical induction shows that:
binary random variables have significant applications in                        (a) Each wk+1 is a linear combination of x1 , . . . , xk+1
modern digital communication systems, such as spread                               and in this combination the coefficient of xk+1 is 1,
spectrum (SS) communication systems or cryptosystems                               hence not 0;
[13, 14]. There is an immense surge in generating the                          (b) Each wk+1 = 0 and so the denominator in the defi-
pseudorandom sequences by chaotic nonlinear maps. Many                             nition of uk+1 is not 0, i.e. uk+1 is well defined and
researchers have proposed different techniques of gener-                            uk+1 = 1.
ating pseudorandom sequences for the secure applications                 (iii)
[13–15, 19]. Khoda and Fujisaki [13] have shown that Ja-
cobian elliptic Chebyshev rational map is a good candi-                         (w2 , u1 ) = (x2 − (x2 , u1 )u1 , u1 )
date in generating the true i.i.d pseudorandom sequences.                                 = (x2 , u1 ) − (x2 , u1 )(u1 , u1 )
This map has many properties e.g. absolutely continuous in-
variant (ACI) measure property, equi-distributivity property,                             = (x2 , u1 ) − (x2 , u1 )1 = 0                   (7)
                                                                                                                     M.K. Khan et al.

Fig. 2 Sample face images
from ORL Database

More generally, if we assume that:                                     2.5 Template verification

                1, i = j ≤ k
(ui , uj ) =                                                           If a person has to verify himself as a legitimate user, then
                0, i = j, i, j ≤ k                                     the proposed system generates a FaceHash by the scheme
then,                                                                  described as above, and matching between the hashed tem-
                                                                       plates can be done by the following equation:
(wk+1 , uj ) = xk+1 −             (xk+1 , ui )ui , uj                       1

                            i=1                                        M=             bi ⊕ bi                                   (10)
               = (xk+1 , uj ) −         (xk+1 , uj )(ui , uj )         where, N is the size of the FaceHash template, bi is
                                  i=1                                  FaceHash-ed template stored in the database, while bi is
               = (xk+1 , uj ) − (xk+1 , uj )(uj , uj )                 a newly generated template for verification.

               = (xk+1 , uj ) − (xk+1 , uj )1 = 0

                                                         def           3 Experimental results
      Hence, (uk+1 , uj ) = 0 if j ≤ k and X = {u1 , u2 , u3 ,
. . . , un } is an orthonormal set of vectors.                         In this section, we provide experimental results and dis-
                                                                       cussions on the proposed scheme. We evaluate the per-
2.4 Inner product and thresholding                                     formance of presented method by using the ORL face
                                                                       database [9]. The database consists of 400 images ac-
Now, we perform the inner product of the X = {u1 , u2 , u3 ,           quired from 40 persons with variations in facial expressions
. . . , un } orthonormal vector set with the normalized face fea-      (e.g. open/close eyes, smiling/non-smiling), and facial de-
ture vector i.e. α by the following equation:                          tails (e.g. with wearing glasses/without wearing glasses).
                                                                       All images were taken under a dark background with a
χ = ( α, u1 , α, u2 , . . . , α, un )                            (8)   92 × 112 pixels resolution. Figure 2 shows two individ-
                                                                       ual samples in ORL database. In the following subsec-
After that, we perform the threshold of vector χ by the fol-
                                                                       tions, we perform experimentation and discuss their re-
lowing equation:
        1      α, ui > λ
bi =                                                             (9)   3.1 Biometric revocation
        0      α, ui ≤ λ

where bi is a revocable biometrics-face Hash, and λ is a               The results shown in Fig. 3 and Fig. 4 indicate that the pro-
threshold value such that on average half bits are zeros and           posed FaceHash scheme is very sensitive and a tiny amount
half bits are ones. At the end, this biometrics FaceHash can           of change in the seed value can abruptly change the be-
be issued to the user and saved in the database or token e.g.          havior of the system. As an example, we used the seed
smart card.                                                            w0 = ABEF28E9CA, but when we modify slightly differ-
Privacy-preserving and tokenless chaotic revocable face authentication scheme

Fig. 3 Key sensitivity
simulation. (a) When seed
w0 = ABEF28E9CA, (b) When
seed w0 = ABEF28E9CB

Fig. 4 Diffusion of FaceHashing at two different seeds. (a) Diffusion when seed w0 = 3839384B4. (b) Diffusion when seed w0 = 3839384B3

ent value of the seed i.e., replace the last alphabet A with B,          3.2 Statistical analysis
w0 = ABEF28E9CB, then the recovered FaceHash is differ-
ent as shown in Fig. 3(b).                                               In order to hide message redundancy, Shannon introduced
    The diffusion of FaceHash at two different seeds is also             diffusion and confusion. These two general principles guide
shown in Fig. 4.                                                         to the design of practical cipher, including Hash functions
                                                                         [15, 18]. For the FaceHash in binary format the ideal diffu-
    Hence, our proposed FaceHash is very sensitive to a
                                                                         sion effect should be that any tiny changes in initial condi-
small change in the seed key and without knowing the seed;
                                                                         tions lead to the 50% changing probability of each bit. For
it is very difficult to generate the same FaceHash. Thus, pre-
                                                                         surveying the performance of diffusion and confusion, we
sented scheme solves the non-revocability problem of bio-                performed the statistical test. If a bit in the seed is randomly
metric templates and a user can be easily issued a new Face-             selected and toggled, and a new hash value is generated,
Hash in the case his previous FaceHash is lost, compro-                  then two FaceHash values are compared and the number of
mised, or theft. Furthermore, our scheme allows generating               changed bit is counted as Bi . This kind of test is performed
different FaceHashes for different array of applications to              N times, and the corresponding distribution of Bi is shown
maintain the privacy of an individual [1].                               in Fig. 5, where N = 2048.
                                                                                                                       M.K. Khan et al.

Fig. 5 Distribution of changed
bits Bi

   If the size of hash is 128 bits, then one bit changed in          Table 1 Number of changed bits Bi
the seed concentrates around the ideal changed bit number-                            N = 512            N = 1024            N = 2048
64bit. It shows that the FaceHash has very strong capability
for diffusion and confusion. Statistical analysis is evaluated       Bmin             44                 46                  46
by the following equations [14]:                                     Bmax             81                 81                  80
                                                                     B                63.97              63.94               63.90
Bmin = min({Bi }N ),
                1             denotes the minimum of Bi      (11)
                                                                      B                5.56               5.30                  5.58
Bmax = max({Bi }N ),
                1              denotes the maximum of Bi     (12)    P (%)            49.95              49.95               49.92
                                                                      P (%)            4.32               4.15                  4.36
¯  1
B=             Bi ,    denotes the mean of Bi                (13)
                                                                     schemes did not discuss the statistical analysis of their gen-
             1                    ¯                                  erated Hashes, so it is difficult to judge the trueness e.g. con-
  B=                        (Bi − B)2 ,
           N −1                                                      fusion and diffusion properties of their BioHashes.

       denotes the standard variance of Bi                   (14)    3.3 FAR and FRR
P = (B/128) × 100%,                                         ¯
                                 denotes the probability of B (15)
                                                                     The performance of the proposed system is measured by
           1                                                         false acceptance rate (FAR) and false rejection rate (FRR).
  P=                  (Bi /128 − P )2 ,
           N                                                         FAR is the rate, at which an imposter generates the same
                                                                     FaceHash of a legitimate user. FRR is the rate, at which an
       denotes the standard variance of P                    (16)    authentic user generates a FaceHash other than his original
                                                                     one. FAR and FRR can be written as:
Through the tests with N = 512, 1024, 2048, respectively,
the corresponding data are listed in Table 1.
                                                                     FAR =     ρ(μn |fI )dx;        FRR =        ρ(μn |fA )dx
    Based on the analysis of the data in Table 1, we can draw
the conclusion that the mean changed bit number B and   ¯
the mean changed probability P are both very close to the            where ρ(μ|f ) is the probability distribution function (PDF)
ideal value 64 bit and 50%. While B and P are very                   of the face features vector f = [f1 , f2 , . . . , fn ], n is the
little, which indicates the capability for diffusion and con-        number of features, and μn is a given class of n = 1,
fusion of our FaceHash is very stable. Besides, all previous         2, . . . , N, N + 1. By computing the following likelihood ra-
Privacy-preserving and tokenless chaotic revocable face authentication scheme

tio, we compute acceptance or rejection strategy [16]:                   References
                    ρ(μn |f )
ρ(μn ) = log                                                   (17)       1. Jain, A. K., Ross, A., & Prabhakar, S. (2004). An introduction to
                   ρ(μN +1 |f )                                              biometric recognition. IEEE Transactions on Circuits and Systems
                                                                             for Video Technology, 14, 4–20.
where μN +1 is the imposters class. By using logarithm’s                  2. Connie, T., Teoh, T., Goh, M., & Ngo, D. (2004). PalmHashing:
property i.e. log a = log a − log b, (16) becomes:
                                                                             a novel approach for dual-factor authentication. Pattern Analysis
                                                                             Applications, 7, 255–268.
ρ(μn ) = log ρ(μn |f ) − log(ρ(μN +1 |f ))                     (18)       3. Connie, T., Teoh, T., Goh, M., & Ngo, D. (2005). PalmHashing: a
                                                                             novel approach to cancelable biometrics. Information Processing
Now first we compute the following:                                           Letters, 93, 1–5.
                                                                          4. Jin, A.T.B, Ling, D.N.C., & Goh, M. (2004). BioHashing: two fac-
μ =      arg       max ρ(μn )                                  (19)          tor authentication featuring fingerprint data and tokenized random
      μ1 ,...,μN                                                             number. Pattern Recognition, 37, 2245–2255.
                                                                          5. Jin, A.T.B, Ling, D.N.C. (2005). Cancellable biometrics featuring
At the end, acceptance or rejection can be performed by the                  with tokenised random number. Pattern Recognition Letters, 26,
following threshold:                                                         1454–1460.
                                                                          6. Maio, D., & Nanni, L. (2005). MultiHashing human authentica-
        if ρ(μ ) ≥ τ       Accept                                            tion featuring biometrics data and tokenised random number: a
D=                                                             (20)          case study FVC2004. Neurocomputing, 69, 242–249.
        if ρ(μ ) < τ       Otherwise                                      7. Nanni, L., & Lumini, A. (2006). Human, authentication featuring
                                                                             signatures and tokenized random numbers. Neurocomputing, 69,
where τ is a decision threshold, and can be defined accord-                   858–861.
ing to the criticality of the biometric system.                           8. David, N. C. L., Andrew, T. B. J., & Alwyn, G. (2006). Biomet-
                                                                             ric hash: high confidence face recognition. IEEE Transactions on
3.4 Analysis of computational complexity and efficiency                       Circuits and Systems for Video Technology, 16, 771–775.
                                                                          9. ORL (1992). The ORL face database. at the AT&T (Olivetti) Re-
All the previous schemes generate pseudorandom sequences                     search Laboratory. Available online:
by using tokens e.g. USB, which is a very costly solution.               10. Turk, M. A. (1991). Eignefaces for recognition. Journal of Cogni-
Generally, if USB connectors are embedded in biometric                       tive Neuroscience, 3, 71–86.
sensors, it would increase the overall cost and these are not            11. Martinez, A. M., & Kak, A. C. (2001). PCA versus LDA. IEEE
commonly available. Here, we assume that if these schemes                    Transactions on Pattern Analysis and Machine Intelligence, 23,
are to be implemented in some sophisticated applications
                                                                         12. Nara, Y., Ran, J., & Suematsu, Y. (2004). Face recognition using
e.g. Automatic Teller Machine (ATM) then, users cannot use                   improved principal component analysis. In IEEE int. symposium
their USBs to generate pseudorandom numbers because of                       on micrometaronics and human science (pp. 77–82).
the limitations in ATM machines. Besides, it is not a feasi-             13. Kohda, T., & Fujisaki, T. (2001). Jacobian elliptic Chebyshev ra-
ble solution to issue each user a specific USB to generate                    tional maps. Physics D, 148, 242–254.
                                                                         14. Khan, M.K, Zhang, J., & Lei, T. (2007). Chaotic secure content-
pseudorandom numbers in performing banking transactions                      based hidden transmission of biometrics templates. Chaos, Soli-
on the ATMs. In our scheme, a user does not need to possess                  tons, and Fractals, 32, 1749–1759.
tokens, e.g. USB, to generate the pseudorandom sequences                 15. Xiao, D., Xiaofeng, L., & Shaojiang, D. (2005). One-way Hash
because the sequence generator is built-in the system and all                function construction based on the chaotic map with changeable-
                                                                             parameter. Chaos, Solitons and Fractals, 24, 65–71.
the processing is done by the presented algorithm.                       16. Erzin, E., Yemez, Y., & Tekalp, A. M. (2006). Multimodal person
                                                                             recognition for human-vehicle interaction. IEEE Multimedia, 18-
4 Conclusion                                                             17. Adams, K., Cheung, K.H., Zhang, D., Mohamed, K., & Jane,
                                                                             Y. (2006). An analysis of BioHashing and its invariants. Pattern
In this paper, we have presented a novel approach of gen-                    Recognition, 39, 1359–1968.
                                                                         18. Khan, M.K, Zhang, J. (2008). Chaotic hash-based fingerprint
erating FaceHashes to maintain the privacy of a biometric                    biometric remote user authentication scheme on mobile devices.
user. Firstly, we pointed out pitfalls and weaknesses of the                 Chaos, Solitons and Fractals, 35, 519–524.
previous schemes, e.g. use of USB to generate pseudoran-                 19. Kia, F., Reza, R., & Hossein, K. (2008). An application of Chen
dom numbers, and then, we proposed an efficient solution                      system for secure chaotic communication based on extended
                                                                             Kalman filter and multi-shift cipher algorithm. Communications
to overcome the irrevocability problem of face-biometric
                                                                             in Nonlinear Science and Numerical Simulation, 13, 763–781.
data by performing simple operations to generate FaceHash
without using the USB token. Experimental and simulation
results have shown that the presented approach is secure,
robust, computationally feasible, and cost-ineffective to at-
tain the revocable face-biometrics authentication. Our future
work will focus on the multimodal BioHashing schemes and
their applications in the real environment.
                                                                                                                                M.K. Khan et al.

                                   Muhammad Khurram Khan is                                                 Khaled Alghathbar Khaled Al-
                                   currently working as Assistant Pro-                                      ghathbar, Ph.D., CISSP, CISM, PMP,
                                   fessor at Center of Excellence in                                        MCSE: Security, Security+, BS7799
                                   Information Assurance (CoEIA),                                           Lead Auditor, is an associate pro-
                                   King Saud University, Saudi Ara-                                         fessor and the director of the Center
                                   bia. He is the Founding Editor of                                        of Excellence in Information As-
                                   ‘Bahria University Journal of In-                                        surance in King Saud University,
                                   formation & Communication Tech-                                          Riyadh, Saudi Arabia. He is a se-
                                   nology (BUJICT)’. He also plays                                          curity advisor for several govern-
                                   role of Editor of several interna-                                       ment agencies. His main research
                                   tional journals of Elsevier Science                                      interest is in information security
                                   and Springer-Verlag. He has been                                         management, policies and design.
                                   the Program Chair and Publica-                                           He received his Ph.D. in Informa-
                                   tion Chair of 12th IEEE Interna-                                         tion Technology from George Ma-
                                   tional Multitopic Conference (IN-                                        son University, USA.
MIC’08). He has also been the Program Chair of the IEEE International
Symposium on Biometrics & Security Technologies (ISBAST’08).                                                Jiashu Zhang received the B.S. de-
He has worked as General Chair for the International Workshop on                                            gree in electronic engineering from
Frontiers of Information Assurance and Security (FIAS’09), Aus-                                             the University of electronic science
tralia. Furthermore, he performed duties of Publicity Co-Chair of 6th                                       and technology of China, Chengdu,
International Conference on Intelligent Computing (ICIC’10), Pub-                                           P.R. China in 1987, and the M.S. de-
licity Co-Chair of 5th International Conference on Intelligent Com-                                         gree in biomedical engineering and
puting (ICIC’09), International Conference on Security Technology                                           instruments from Chongqing Uni-
(SecTech’09), International Conference on Ubiquitous Computing and                                          versity in 1990, and the Ph.D. de-
Multimedia Applications (UCMA’10). He is an advisory board of                                               gree in communication and infor-
2nd International Conference on Advanced Science and Technology                                             mation system form University of
(AST’10). Moreover, he is workshop management chair of 4th Inter-                                           electronic science and technology
national Conference on Information Security and Assurance (ISA’10).                                         of China, Chengdu, P.R. China in
He also works as the Program Committee Member of ICPR’10,                                                   2001. He is currently a full profes-
ICWAPR’10, AICCSA’10, ICIC’10, PDCAT’09, ACSA’09, RISC’09,                                                  sor of information and communica-
ICWAPR’09, ScalCom’09, ICIC’09, ICIC’08, ICPADS’08, HPCC’08,                                                tion engineering in the school of in-
ICWAPR’08, ICIC’07, and ICIC’06. Besides, he is a reviewer of sev-        formation science and technology at Southwest Jiaotong University,
eral international journals and conferences. Recently, he has been        Chengdu, P.R. China. His current research interests are in the areas of
awarded outstanding leadership award at IEEE NSS’09 conference            biometric and information security, signal processing for communica-
in October, 2009 at Australia. He has been recently included in the       tion, nonlinear system and chaos.
Marquis Who’s Who in the World 2010 edition. Dr. Khurram has pub-
lished more than 50 research papers in the journals and conferences of
international repute. His areas of interest are biometrics, information
security, multimedia security, and digital data hiding.

Shared By: