Adaptive Threshold for Fingerprint Recognition System Based on Threat Level and System Load
Vaibhav B. Joshi, Mehul S. Raval · Procedia Computer Science · 2020
A fingerprint is a widely used trait for person recognition in civilian applications. User authentication takes place when the matching score is higher than the acceptance threshold. The performance of the fingerprint system (FS) is evaluated based on a false acceptance rate (FAR) and false rejection rate (FRR). Usually, the FS is set to work at a rate where FAR and FRR are equal (EER). However, operating at EER allows finite FAR, which is risky during the critical threat. In response, the acceptance threshold must shifts towards zero FAR to mitigate the threat. It increases FRR, system load, and user inconvenience. In the civilian application acceptance threshold is set by the vendor, and currently, there is no research attempt to change it dynamically. It is necessary as; 1) system must respond to external parameters like load and threat level; 2) system must balance security and user convenience due to high traffic. This paper describes a method to change the acceptance threshold over the interval EER to zero FAR based on system load and threat level. The proposed method uses a fuzzy inference system (FIS) and an artificial neural network (ANN).