Research On Biometric Cryptographic Key Using Electrocardiogram And Fingerprints

As for the human-related information security applications, such as non-repudiation accountability and biomedical information confidentiality, traditional "device-centered" key management mechanisms can neither satisfy the new requirements of security laws and standards, nor be suitable for the new energy-efficient constraints in the power-limited environments. Therefore, this paper combines the biometrics with cryptography, and investigates "human-centered" biometric key protection and biometric key distribution techniques. Since the biometric measurements are usually noisy and fuzzy, and cannot be revoked or canceled if compromised, the biometric cryptographic key techniques should be able to rise to lots of new challenges: (i)Security problem: how to provide the security properties of the biometric privacy and the secret keys; (ii)Accuracy performance: how to deal with the contradiction between the noisy biometric information and the exactness of the cryptographic keys; (iii)Length of secured template: how to reduce the size of the communication data and the energy consumption for biometric-based key distribution.The work is summarized as follows:1) As for the deterministic extraction process of fuzzy and noisy information, a series of new threshold-based extraction algorithms, FuzzyLock scheme based on the characteristic polynomial reconstruction, are proposed. This paper further investigates two biometric cryptographic key algorithms based on set difference threshold and set intersection threshold respectively, and provides the systematic security analysis. Moreover, when one biometric is used for multiple applications, a new reusable fuzzy cryptographic technique, Randomized Dissolvent Template (RDT), is proposed to deal with the multiple templates attack, which is one of the most difficult attacks for the current biometric cryptographic key algorithms.2) Since traditional biometric feature extraction and recognition approaches cannot be applied to the biometric cryptographic key algorithms directly, this paper proposes new schemes that can represent the biometrics as integer sets with fixed-value elements, including VitaCode-based electrocardiograph (ECG) biometric representation, and MinuCode-based fingerprint biometric representation approaches, so as to serve as a suitable and accurate data basis for biometric cryptographic algorithms.3) A novel energy-efficient biometric key distribution approach, BodyKey, is proposed, which integrates the 1st kind of FuzzyLock algorithm with the ECG biometrics. The experimental results on the PysioBank ECG database show that: BodyKey can deal with the fuzzy noise of the ECG features and provide better accuracy performance; moreover, BodyKey is able to reduce the length of the secured template and further achieve lower energy consumptions for key distribution in the body area networks.4) A new fingerprint-based biometric key protection approach, MinuKey, is proposed, which integrates the 2nd kind of FuzzyLock algorithm with the fingerprint biometrics. The experimental results on the FVC2002-DB fingerprint database show that: MinuKey is able to handle the fuzzy noise of the fingerprint features and provide better accuracy performance effectively, and provide the merits of security properties of key protection, biometric privacy, key non-inversion and biometric reusability, and consequently establish trustworthy connection between the logical cryptographic key and the physical legitimate user.