A Fingerprint Matching Model using Unsupervised Learning Approach

Nasser Abouzakhar, Muhammed Bello Abdulazeez

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    183 Downloads (Pure)


    The increase in the number of interconnected information systems and networks to the Internet has led to an increase in different security threats and violations such as unauthorised remote access. The existing network technologies and
    communication protocols are not well designed to deal with such problems. The
    recent explosive development in the Internet allowed unwelcomed visitors to gain
    access to private information and various resources such as financial institutions,
    hospitals, airports ... etc. Those resources comprise critical-mission systems and
    information which rely on certain techniques to achieve effective security. With the
    increasing use of IT technologies for managing information, there is a need for
    stronger authentication mechanisms such as biometrics which is expected to take
    over many of traditional authentication and identification solutions. Providing
    appropriate authentication and identification mechanisms such as biometrics not
    only ensures that the right users have access to resources and giving them the right privileges, but enables cybercrime forensics specialists to gather useful evidence whenever needed. Also, critical-mission resources and applications require mechanisms to detect when legitimate users try to misuse their privileges; certainly biometrics helps to provide such services. This paper investigates the field of biometrics as one of the recent developed mechanisms for user authentication and evidence gathering despite its limitations. A biometric-based solution model is proposed using various statistical-based unsupervised learning approaches for fingerprint matching. The proposed matching algorithm is based on three various similarity measures, Cosine similarity measure, Manhattan distance measure and Chebyshev distance measure. In this paper, we introduce a model which uses those similarity measures to compute a fingerprint’s matching factor. The calculated matching factor is based on a certain threshold value which could be used by a forensic specialist for deciding whether a suspicious user is actually the person who claims to be or not. A freely available fingerprint biometric SDK has been used to develop and implement the suggested algorithm. The major findings of the experiments showed promising and interesting results in terms of the performance of all the proposed similarity measures.
    Original languageEnglish
    Title of host publicationProcs 3rd International Conference on Cybercrime Forensics Education & Training
    Subtitle of host publicationCFET 2009
    Number of pages12
    Publication statusPublished - 1 Sept 2009
    Event3rd International Conference on Cybercrime Forensics Education & Training - Canterbury, United Kingdom
    Duration: 1 Sept 20092 Sept 2009


    Conference3rd International Conference on Cybercrime Forensics Education & Training
    Country/TerritoryUnited Kingdom


    • Biometric security
    • Fingerprint biometrics
    • Unsupervised learning


    Dive into the research topics of 'A Fingerprint Matching Model using Unsupervised Learning Approach'. Together they form a unique fingerprint.

    Cite this