Electrical Engineering Department
PhD Thesis Defense Session | | |
Robust Combination Methods for Biometric Systems |
Abstract: Designing multi modal biometric systems proposes more precise and confident technique than single systems, with one biometric modality and one matcher, for authentication. Because absolute security is impossible, investigation of various invasions to these systems and analyzing vulnerability of systems against these attacks cause proposing suitable methods for designing secure authentication systems and increment their robustness. This thesis introduces the possibility and probability of spoof attacks to each modality of biometric matchers and investigates the vulnerability of combined biometric systems against spoofing of their matchers' modalities. In this research for case study face and fingerprint traits are fused in score level for generation of bimodal biometric systems. For studying the vulnerability of systems, we have simulated spoofed scores of each matcher, up on proposed parameters by a novel method, then imposed invasion to authentication systems by these simulated scores, and at last calculated the vulnerability of them by an applicable method. Vulnerability of biometric systems is evaluated in individual and bimodal forms by the most common fusion rules, included fixed combination rules contained analytic and fuzzy rules, and linear and nonlinear trainable rules against various possibilities and probabilities of spoofing and experimental results have been analyzed and compared. Comparison of quality and quantity outcomes of various fusion rules shows the robustness and vulnerability of each of them. In this thesis, spoofed scores database has been constructed from real standard scores database with various possibilities and probabilities of spoofing. Following the goal of research and for improve the robustness of biometric systems against spoofing, rules and guidelines include: finding suitable biometric modality and its matcher, robust combination rule and its robustness conditions are proposed as the contribution of these research. Novel applied experimental researches and extraction method of their results show the importance of consideration to and more investigation the robustness of biometric systems and can help to development of security and creation the appropriate robustness in biometric systems by investigation their vulnerabilities with fusion rules and biometric traits.
|
Mahdi Hariri Supervisor: Dr Shahriar Baradaran Shokouhi Advisor :Dr Sattar Mirzakuchaki The Jury:Dr Ehsanollah Kabir, Dr Nasrollah Moghaddam, Dr Karim Mohammadi, Dr Ahmad Ayatollahi, Dr Golam ali Rezaei Rad
|
Viva Date: Sunday 8 Jan 2012 Place: Electrical Engineering faculty, Seminar Room (Class 303)
|