PClusBA: A Novel Partition Clustering-Based Biometric Authentication Mechanism

In any communication paradigm, authentication plays an important role. Common authentication methods such as passwords are vulnerable to brute force attacks and cryptanalysis. Authentication mechanisms based on biometrics prove to be a better alternative owing to their permanence, uniqueness, and non-reproducibility. Fingerprint biometrics is more suitable in the context of network security applications due to its uniqueness and ease of processing. In general, each fingerprint has 30-40 unique points. The lesser the number of unique points extracted from the fingerprint, the more flexible it is to represent a Unique ID (UID). Clustering can be used to prune the unique point count. Since partition clustering gives us the choice of selecting the cluster count apriori, we have used partition clustering in the proposed biometric authentication mechanism named PClusBA. The cluster count is set to eight to obtain a UID of 192 bits in the proposed format. The length of the UID conforms to the NIST standards for enhancing security in commercial applications. The algorithm was implemented and tested using images from DB1 FVC 2002 database and the results were found to be unique for each image.

Automated summary

The text discusses an authentication mechanism based on fingerprint biometrics, optimizing the extraction of unique points using partition clustering for higher uniqueness and security. The proposed mechanism, PClusBA, uses a cluster count of eight to achieve a 192-bit UID in accordance with NIST standards, and was found to provide unique results for each image tested.