A biometric encryption system algorithm development and system level design

Abstract

Biometric encryption (BE) is a security scheme enhancement that overcomes the exploitable vulnerabilities of biometric authentication systems and the key storage issues of cryptographic schemes by combining both those systems. The practical application of security schemes often requires them to be stand-alone devices with tamper-resistant hardware implementation such as a System-on-Chip (SoC). Therefore, it is suitable to design a BE system architecture that is enhanced for speed and performance in a resource-constrained environment. This thesis proposes a novel algorithm for chaff generation (CG), which is a highly computeintensive algorithm in the BE system. The proposed CG algorithm is suitable for hardware implementation in an SoC because it is proven to have lower algorithmic complexity of O(n2) compared to the existing Clancy’s CG algorithm that has O(n3) complexity. Experimental results have shown that the proposed algorithm is about 150 times faster. Furthermore the proposed CG algorithm overcomes the security vulnerability detected in Clancy’s CG algorithm. The design of such a complex system, which contains many compute-intensive algorithmic blocks, requires consideration of multiple options for system architecture, optimal hardwaresoftware partitioning and early design verification. Hence, state-of-the-art systemlevel modeling using SystemC is applied in the design of the proposed BE system. In this thesis, the system-level design process has been enhanced by adding the Algorithmic Model level on top of the existing design abstraction levels. To verify the functionality of the BE design, appropriate testbenches must be generated and refined along with the system model throughout the different design abstraction levels. For this purpose, a new verification framework with a testbench generation methodology is also proposed, which generates testbench at the algorithmic level using MATLAB and incrementally refines it for use at lower levels of the design abstraction. This framework is applied in the early system-level verification of the proposed BE system. Experiments conducted have also shown that the proposed verification framework that integrates MATLAB testbenches with SystemC facilitates the verification process and reduces verification time through testbench refinement.