HIERARCHICAL MINUTIAE MATCHING FOR FINGERPRINT- BASED CARDLESS AUTOMATED TELLER MACHINE (ATM)

Abstract

Automated Teller Machine (ATM) is an electronic banking outlet, self-service technology and financial service delivery which allow banks customers to seamlessly perform transactions without the aid of bank officials and usually adopted by financial institutions to reach their customers outside/inside the banking hall. The user of existing ATM uses ATM card to access their account to perform one or more financial transactions. Several problems are associated with the use of ATM card such as card cloning, card damaging, card expiring, cast skimming, cost of issuance and maintenance, accessing customer account by third parties, waiting time before issuance expiring or new card. With these problems, the use of ATM card has become a treat to safety of customer funds, even though the stakeholders in financial transactions are making great efforts to reduce ATM frauds. This Thesis presented the design and implementation of a conceptual framework of cardless ATM that uses fingerprint technology to control access to the ATM using Hierarchical PCA technique. The summed-up Hough algorithm was utilised for coordinating system parameters by considering two particular sets P and Q where P is one of the database layouts unique mark sets, and Q is the question unique mark format's details set. To match these two sets of points together, the space of all possible transformations of the point-set are defined by the following transformation parameters: scale (s), rotation (φ), and Cartesian displacement (Δx and Δy). The simulation model was mapped into the system by studying the current system’ operation and another layer of authentication (fingerprint) was added to it. Finally, the proposed framework was implemented using C#, a .NET programming language environment. The system performance evaluation was carried out using Accuracy, Precision, Sensitivity and True Negative Rate metrics. Out of 299 samples, there were 299 (98.66%) correct classifications containing 112 for Complete, 167 for Incomplete and 16 for Out of service (along the diagonal) and 4 (1.34%) incorrect classifications containing 2 Complete as Incomplete and 2 out of service as Incomplete. From the information provided by the confusion matrices, the system accuracy is 98.66%, and the precision is 0.9923.