Thesis presentation

Presented By: Kareem Kamal A. Ghany Under Supervision of: Prof. Aboul Ella Hassanien Prof. Hesham A. Hefny

4, June 2014

INTRODUCTION

Problem Definition

Research Objective

Thesis Structure

OVERVIEW

Biometrics

Machine Learning

Related Work

PROPOSED APPROACH

A Symmetric Bio-Hash Function Based on Fingerprint

Minutiae and Principal Curves Approach

Hybrid Approach for Protecting Biometrics Template

Watermarking effect on the Biometric System

CONCLUSION

FUTURE WORK Kareem Kamal A.Ghany 2

The main problem of this research work is to

find a general way to make a hybrid

biometric system based on machine learning

techniques more effective in a broader range

of applications for increasing performance of

security systems.

3 June 2014 Kareem Kamal A.Ghany 3

Provide an intelligent hybrid

biometrics system that uses a

cryptosystem approach combined

with a transformation based

biometric approach

Prove that the watermarking

technology does not affect the

performance of the biometric

system


To increase the performance of the security system.

To decrease the error rate in the biometric system.


Ch2: Biometrics: An overview

Ch3: Machine Learning techniques : An overview

Ch4: Proposed Unimodal Biometric approach

Ch5: Proposed Hybrid Biometric approach

Ch6: Proposed watermarking approach based on DNA and Fingerprint

Ch7: Conclusion & Future work


A biometric is a unique, measurable characteristic or trait of a human being for

automatically recognizing or verifying their identity.


The September 11 Terrorist Attacks

Following the September 11 attacks (2001), The use of biometrics is likely to increase in the future as security concerns become more of a priority for both governments and corporations.


Only biometrics can verify you as you

Tokens (smartcards, etc.) aren't you and can be: lost stolen duplicated (some) forgotten

Passwords aren't you and can be: forgotten shared observed broken

A multi-biometric system can be classified into one of the following six categories:

multi-sensor, multi-algorithm, multi-instance, multi-sample, multi-modal, or hybrid.


Hybrid systems: The term hybrid used to

describe systems that integrate a subset of

the other five scenarios. For example,

discussing an arrangement in which

fingerprint recognition algorithm id

combined with DNA recognition algorithm

at the match score.

Physical biometrics Behavioral biometrics

Examples:

- Speaker recognition.

- Signature recognition.

- Gait recognition.

- Keystroke dynamics.

Examples:

-Fingerprint.

- Iris print.

- Face print.

-DNA print.




- It’s how the computer (machine) would automatically extract the algorithm for any

task .

- The research is aim to illustrate how ML techniques can be applied to solve

biometrics problems and how biometrics data can be analyzed, processed, and

classified by ML.

- Example of ML techniques:

Neural Networks Kernel Methods

Support Vector Machine (SVM) Near Sets

Fuzzy sets Principal Component Analysis (PCA)

Genatic Algorithm (GA) Rough Sets



It is a way of identifying patterns in data,

and expressing the data in

such a way as to highlight their similarities

and differences.

PCA is a powerful tool for analyzing data.

It can be used for feature extraction

purposes in its special case (Principal

Curves Approach)

In (Juels and Sudan, 2002) they proposed a fuzzy vault scheme which embeds a

secret S in a fuzzy vault with a dataset A. In order to extract the secret, one needs to

present another set B to decrypt the vault.

In (Cheung et al., 2005) authors revealed that noninvertible process can be invertible

to the extent that the matcher may be fooled through a case study on bio-hashing and

face.

In (Uludag et al., 2005) they proposed a model to explore the realization of a

cryptographic construct, with the fingerprint minutiae data. This construct aims to

secure critical data such as secret encryption key with the fingerprint data in a way

that only the authorized user can access the secret by providing the valid fingerprint.


Cancelable biometrics was first proposed by (Ratha et al., 2006), An improvement

for their work is presented through accurate registration which is a key step in

building cancelable transforms, the overall approach has been tested using large

databases and the results demonstrate that without losing much accuracy a large

number of cancelable transforms for fingerprints can be built.

In (Feng et al., 2008), they proposed a three-step cancelable framework which is a

hybrid approach for face template protection. This hybrid algorithm is based on the

random projection, class distribution preserving transform and hash function.


In (Kumar and Vikram, 2010) they proposed a new method of recognition of the

pattern of distribution of the minutiae points of an enhanced image, and they used a

multi-dimensional artificial neural network. the advantage of their method is the

usage of the entire resized minutiae image as an input at once. It is capable of

excellent pattern recognition properties as the distribution of the minutiae points are

used directly for recognition.

In (Nagar et al., 2010) they proposed a hybrid approach that improves the matching

performance as well as security of the fuzzy vault, one limitation of a fuzzy

commitment scheme designed using typical algebraic codes is that these codes

do not meet the Hamming bound.

In (Kekre et al., 2010) they proposed image retrieval techniques based on features

extracted from Kekre transform and then applied them on row mean, column mean

and combination.

17

1:Proposed Uni-modal Biometric approach

2:Proposed Hybrid Biometric approach

3:Proposed watermarking approach based on

DNA and Fingerprint





DNA and Fingerprint



Minutiae Types


Same Fingerprints with Different Positions

Securing Fingerprint Template

We use bio-hash function as a Unimodal Biometric System to secure the fingerprint template as showing in this Figure Put Algorithm after this slide

Uni-modal Biometric Approach


Shifting of minutia points

The transformation done from one fingerprint to another can be described by the following equation: Where z represents the minutia point, r represents the minutia rotation, and t represents the minutia translation, this Figure shows an example of shifting the minutia points.

Proposed Uni-modal Approach


Where: r is the rotation.

t is the translation


Large distance between the points

tends to large error rate

To Minimize the error rate we will

apply the Hybrid Approach (next

contribution)

A.GHANY Kareem Kamal, et. al, ”A Symmetric Bio-Hash Function Based On Fingerprint Minutiae

And Principal Curves Approach”, The 3rd International Conference on Mechanical and Electrical

Technology ( ICMET 2011), ASME, China, vol. 1, 2011.




DNA and Fingerprint


Class Distribution Preserving (CDP) is a technique used for template transformation

by using a set of ”distinguish point” and distance measurement. The ”distinguish

point” set is randomly generated, which will increase the randomness of the

representation. CDP transfers a real value template into a binary template. The

main idea is to make use of the group of distinguishing points, a distance function

and the threshold.

Proposed

Hybrid

Approach

Transformation

Approach

Biometric

Cryptosystem

Approach C

DP

Tra

nsf

orm

Sym

met

ric

Has

h

Fu

nct

ion







A.GHANY Kareem Kamal, et. al, ”A Hybrid approach for biometric template security”, The 2012

IEEE/ACM international conference on Advances in Social Network Analysis and Mining

(ASONAM 2012), Turkey, pp. 941-942, 2012.

Kekre’s transform is a technique used for template transformation which include in

its traditional methods the template that transformed using the user’s password

during the enrollment, and the authentication.

Proposed

Hybrid

Approach

Transformation

Approach

Biometric

Cryptosystem

Approach K

ekre

’s

Tra

nsf

orm

Sym

met

ric

Has

h F

un

ctio

n


Our new proposed approach contains four phases: A. Feature extraction using principal curves approach. B. Kekre’s transform. C. Securing the template by using Bio-Hash function. D. Matching Phase.

During authentication, the query is used to recover

the original biometric template from the secure

sketch and the exact recovery of the original

biometric data is verified to authenticate a user.

Also, new hash values are produced by the

scanner and are matched with those stored in the

database.

We find that matching can be performed on both

the client side and on the server side, and it is

performed using hashed features instead of the

original template.

Authentication is depending on scores which can

range between percentage from 0% to 100%.

Hybrid Biometric Approach Kekre’s transform


Kekre’s transform Vs. Class Distribution Preserving (CDP)

CASIA database gave the accurate results of 85% which is higher than CDP, shows

that using Kekre transform is better than CDP transform by 20%.




E in [2] is the error rate when using CDP transform.

E in new approach is the error rate when using Kekre’s transform.



A.GHANY Kareem Kamal, et. al, ”Kekre’s transform for protecting fingerprint template”, The 13th

International Conference on Hybrid Intelligent Systems (HIS 2013), IEEE, Tunisia, pp. 186-191,

2013.




DNA and Fingerprint


Performance measurements of Fingerprint Image and DNA


DNA typing is very useful in crime detection.

Since DNA requires a form of blood, tissue, or other bodily sample, it has not

yet been adopted as a major biometrics methods, even though it is now

possible to analyze human DNA within 10 minutes.


Digital watermarking methods

should be imperceptible in order

to be effective, while at the same

time robust to common image

manipulations like rotation,

compression, scaling, filtering,

cropping, and collusion attacks

through other signal processing

operations.

Digital Image Watermarking

Spatial

Domain

Frequency Domain

DCT DFT DWT


The hierarchical property of the DWT offers the possibility of analyzing a

signal at different resolutions (levels) and orientation. This multi resolutions

analysis gives both space and frequency localization, and different

orientations extract different features of the frame, such as vertical, horizontal

and diagonal information. The DWT of an image has two parts: an

approximation part (this is an image with smaller dimensions) and a detail

part (this is a set of images with smaller dimensions containing the details of

the original image).


Input Image

LL1 HL1

LH1 HH1

HL1

LH1 HH1

HL2

HH2 LH2

LL2

HL1

LH1 HH1

HL2

HH2 LH2

LL3 HL3

HH3 LH3


DNA Encoding

Example:

If we have the sequence T10101101U It will be converted to

TTTGAU



Our new proposed approach contains four phases: A. DNA Encoding. B. Watermarking Embedding Using DWT. C. Watermark Extraction Using DWT. D. DNA Decoding.


Proposed watermarking approach based on DNA and Fingerprint Algorithms

The experimental results for the proposed approach have been

evaluated using the following measures:

1- Structural Similarity Index Measure (SSIM):

It designed for quality assessment, it compares local patterns of pixel intensities that

have been normalized for luminance and contrast.


2- Peak Signal to Noise Ratio (PSNR):

It is a mathematical measure of image quality based on the pixel difference between

two images, it’s an estimate of quality of reconstructed image compared with

original image..


3- Mean Square Error (MSE):

It is computed by averaging the squared intensity of the original (input)

image and the watermarked (output) image pixels.

Where I and W are the original and the watermarked images having a resolution

of (m * n)


Histograms of original and watermarked image





A.GHANY Kareem Kamal, et. al, “A Hybrid Biometric Approach Embedding DNA

Data in Fingerprint Images”, The 3rd IEEE International Conference on Informatics,

Electronics & Vision (ICIEV), Bangladesh, May 23-24, 2014.

This thesis addressed the problem of securing Biometric system, therefore, we

applied a Bio-hash function only then combined it with both the Class Distribution

Preserving (CDP) and the Kekre’s transform to compare the effect of them on the

error rate, i.e. on the performance of the biometric system.

In This thesis we also applied a watermarking technique on a hybrid biometric

system using fingerprint and DNA, and we prove that the watermarking technique

doesn’t affect either the quality of the biometric image or the performance of the

biometric system.


In the future work we need to apply our approach with new technologies, such as:

1- Apply the Biometric System on Cloud Computing.

2- Apply the Hybrid approach to secure the data on Cloud.

3- Try Biometrics of Animals rather than Human.



1- A.GHANY Kareem Kamal, A.Moneim Mahmood, Ghali Neveen I., Hassanien

Aboul Ella, Hefny Hesham A., "A Symmetric Bio-Hash Function Based On

Fingerprint Minutiae And Principal Curves Approach", The 3rd International

Conference on Mechanical and Electrical Technology ( ICMET 2011),ASME,

China, vol. 1, 2011.

2- A.GHANY Kareem Kamal, Hefny Hesham A., Hassanien Aboul Ella, Ghali

Neveen I., "A Hybrid approach for biometric template security", The 2012

IEEE/ACM international conference on Advances in Social Network Analysis and

Mining (ASONAM 2012), Turkey, pp. 941-942, 2012.

3- A.GHANY Kareem Kamal, Hefny Hesham A., Hassanien Aboul Ella, Tolba M.

F., "Kekre's transform for protecting fingerprint template", The 13th IEEE

International Conference on Hybrid Intelligent Systems (HIS 2013), pp. 186-191,

2013.


4- Kareem Kamal A.Ghany, Aboul Ella Hassanien and Gerald Schaefer “Similarity

Measures for Fingerprint Matching” Int. Conference on Image Processing, Computer

Vision, and Pattern Recognition, 21-24 July, 2014, Las Vegas, USA, 2014.

5- Kareem Kamal A.Ghany, Gehad Hassan, Gerald Schaefer, Aboul Ella Hassanien

Md. Atiqur Rahman Ahad, Hesham A. Hefny “A Hybrid Biometric Approach

Embedding DNA Data in Fingerprint Images” The 3rd Intl. Conf. on Informatics,

Electronics & Vision. Dhaka - Bangladesh, 23-24 May, 2014.

6- A.GHANY Kareem Kamal, Hassanien Aboul Ella, Hefny Hesham A., "Machine

Learning in Biometrics: A Review", Book Chapter, Springer, 2014, (Submitted).

7- A.GHANY Kareem Kamal, Sayed Gehad I., Anwar Asmaa S., Ibrahim Fatma M.,

Fathy Ghada D., Hassanien Aboul Ella, Hefny Hesham A., "Securing Maps Data

Using Hybrid Biometrics Watermarking Scheme", International Journal of Coding,

Designs, and Cryptography, Springer, 2014, (Submitted).


http://www.egyptscience.net/

http://www.egyptscience.net/


Thesis presentation

Documents

biometrics data

hybrid biometric system

hybrid algorithm

cancelable biometrics

watermarking approach

multibiometric system

fingerprint data

related work proposed