A Study Of Fingerprint Recognition And Encryption Based On Entropy Analysis

The biometric security is paid much attention to the researchers accompanied bythe growth of biometric technology application in the information security field inrecent years. The research on biometric characteristics representation and biometricencryption are particularly important. Entropy, a basis of information theory, is a crucialindex for the information processing and security analysis in cryptography. It is ameasure which reflects the amount of information of the signal. Entropy also can beused as a description of the information randomness. The entropy loss shows us thechanges in the randomness of information. The entropy analysis conduces us tounderstand the relationship between biometric and encryption technology for it. In thispaper, the entropy analysis is used for two specific issues. The first one is the singularpoint detection in the fingerprint foreground region, where the multi-scale orientationentropy is as a representation for the singularity of fingerprint. The second one isconcerned with the security analysis of a fingerprint encryption method for messagedeadlock problem, where multiple minutia sets are used with the same privilegesrelative to the decoding procedure. The author’s major contributions are outlined asfollows:1. A singularity representation and fingerprint singular point detection method havebeen studied. Firstly, according to the difference of orientation entropy between in theconsistent ridge-valley region and singularity region in the fingerprint foreground, aentropy analysis based method is proposed for singular point detection. By analyzingthe orientation entropy with the multiple scale analysis, it is found that the informationentropy is higher in the fingerprint singularity region than the consistent ridge-valleyregion. Some post processing steps are proposed to filter the spurious core and deltapoints based on the heuristic rules of singular point. Although the smooth orientation ishelpful to reduce noise impact against the calculation of orientation entropy, an iterationmethod is proposed to compensate the slight offset. The proposed method is based onthe coarse orientation flow curve extraction. Experimental results show that themultiple-scale orientation entropy is correct and effective for singular detection. Thecompensation method reduces the offset between the detection result and truth singularpoint.2. The message deadlock problem in the biometric encryption system has beenstudied. Using the modified fuzzy vault scheme, in which the chaff points are filtered according to the match method based on the minutia neighbor structure, a secret sharingscheme based on multiple biometric instances is designed. For all participants in thescheme, they share the secret with equally privileges. Compared with the other methodsfor the same issue, a multi-vault structure is constructed in the proposed method. Itsimplifies the selection step of truth unlocking set in the single vault structure. Thesecurity analysis of the proposed method is given by means of min entropy and entropyloss. Experimental results show that the proposed scheme decreases the deadlock ratewithout increasing the false acceptance rate significantly.3. An algorithm platform for the biometric recognition and encryption has beendesigned by using the Emgu package, where the platform is based on the Windowsclasses for the graphical display. In the platform, some fingerprint recognitionalgorithms have been coded such as orientation extraction, fingerprint foregroundsegmentation. It is expected to be helpful for the further research.