1. Multimodal Biometric Security
Presented By: Qurratulain
PhD Scholar
SEIEE Department,
Shanghai Jiao Tong University

2. Introduction
An application of modern statistical methods to the measurements of biological objects
PIN
Identification of an individual on the basis of “Who you are?” rather than
“What do you know?”
Knowledge-based security, e.g. a password , a PIN..
“What do you have?”
Token-based security, e.g. passport, driver’s license, ID card..
Know Be
Have

3. Introduction
Working of a biometric system
Test
Test

4. An example of multibiometric system
Introduction
Sensors
Extractors
Image- and
signal- pro.
algo.
Classifiers
Negotiator
Threshold
Biometrics
Voice, signature
acoustics, face,
fingerprint, iris,
hand geometry, etc
Data Rep.
1D (wav),
2D (bmp,
tiff, png)
Feature
Vectors
Scores
Enrolment
Training
Submission
An example of multibiometric system

5. Introduction Need for Multimodal Biometrics Limitations
Increase security
Make it possible, automatically, to know WHO did WHAT, WHERE and WHEN!
Replace hard-to-remember passwords
Prevent unauthorized use of lost, stolen or "borrowed" ID cards
Limitations
Noise in sensed data results in a significant reduction in accuracy
For voice and signature, circumvention by spoofed attacks

6. Problem Statement
The use of signature verification is a time-consuming process for the user in real time situations
User-involvement at image acquisition stage reduces the time efficiency of the identification process and causes delay
Mostly PCA and LDA based global features are used for feature extraction and superiority of LDA over PCA has been affirmed many times so far, therefore a newer approach for feature extraction is required that can outperform the existing methods e.g. Discrete Cosine Transform (DCT) because of its energy compaction property

7. Problem Statement
Moreover, security concerns have been increased because of the following drawbacks of unimodal biometric systems:
Noise in sensed data results in a significant reduction in accuracy
Non-universality leads to Failure To Enroll error (FTE)
Lack of Individuality or Uniqueness increases False Accept Rate (FAR)
Openness to circumvention by spoofed attacks commonly for voice and signature

8. Solution
A multibiometric system can be used based on feature Level fusion and feature vector optimization through genetic algorithm.
Multibiometric system can be implemented so that it can reduce error rate caused by unibiometric systems, e.g.,
If the biometric sample obtained from one of the sources is not of sufficient quality during a particular acquisition, the samples from other sources may still provide sufficient discriminatory information to enable reliable decision-making.
Non-Universality
If a person cannot be enrolled in a finger print system due to worn-out ridge details, he can still be identified using other biometric traits like face or iris
Noisy data

9. Solution
Discrete Cosine Transform (DCT) can be employed along with the old techniques of feature extraction like PCA and LDA
DCT has been selected as it offers many benefits:
Provides good compromise between energy packing ability and computational complexity
The energy packing property of DCT is superior to that of any other unitary transform
Transforms that redistribute or pack the most information into the fewest coefficients
Provides the best sub-image approximations and, consequently, the smallest reconstruction errors

10. Solution
A multimodal biometric system can be introduced that reduces user involvement at the image acquisition stage
Discrete Cosine Transform (DCT) can be employed along with the old techniques of feature extraction like PCA and LDA
Normalization of feature vector can be performed using z-score
Normalized feature vectors can be fused to form multibiometric system, fusion being done by product, sum and min rules
Genetic Algorithm (GA) can be used for feature vector optimization after feature level fusion
Selected features can then be incorporated into a Bayesian Classifier for calculating the accuracy of the proposed system

11. Proposed Solution Face image Iris image Ear image Face Preprocessing
Iris Preprocessing
Ear Preprocessing
Face, Iris, and Ear Features
DCT System
Face, Iris, and Ear Features
PCA System
Face, Iris, and Ear Features
LDA System
Feature Vector Normalization
Feature Vector Normalization
Feature Vector Normalization
Feature level Fusion
Optimization using Genetic Algorithm
Bayesian Classifier

12. Future Work
In the field of Multibiometrics, there is still a lot of work requied to reduce the effects of noisy data and to increase template security
The work presented in this thesis can be extended:
By implementing different variation of PCA and LDA like Multilinear PCA, kernel PCA, Independent Discriminant Analysis, etc
By using the energy compaction property of DCT on a feature extraction method of LDA, and PCA. In this way the feature vector of LDA and PCA systems would be optimized by DCT
By employing various fusion techniques in parallel and concatenating the resultant vectors.
The different levels of fusion can be evaluated for making a comparison on the multibiometric system

13. Thank you