A Data Mining–Based Multimodal Biometric Authentication System for Secure Access Control

Abstract

The rising need to have safe and trustworthy access control systems has increased the rate at which biometric authentication systems are being adopted. Unimodal biometric systems are however usually plagued with limitations like noise, intra-class variations and spoofing attacks. In order to overcome these obstacles, this paper presents the suggested multimodal biometric authentication system based on data mining; the system combines fingerprint, face, and iris modalities through feature-level and score-level fusion. The evaluation of the system performance based on the conventional measures of biometrics was an experiment based on the following measures: accuracy of authentication, False Acceptance Rate (FAR), False Rejection Rate (FRR), system reliability, and acceptance of the system by its users. Results of the experiment indicate that the suggested multimodal system has a great accuracy of authentication of 98.2% with much lower FAR (0.9) and FRR (1.3) than unimodal systems. Moreover, the system is highly reliable (96.9%) and is well accepted by the users (95.4%), which proves its strength and practicability. The results indicate that the pattern of using a combination of several biometric characteristics plus data mining methods are effective in increasing the security, accuracy, and general system efficiency, which implies the framework of correlation to the actual secure access control systems.