<![CDATA[Access control systems are essential tools for combating the nefarious actions of malicious actors in the digital space. Multimodal biometric access control systems are considered state-of-the-art in access control systems, however, existing approaches suffer from limited classification performance. This study integrates face, ear, and iris recognition to develop a trimodal Chicken Swarm Optimization (CSO)---enhanced Self-Organizing Feature Map (SOFM) classifier. Six high-resolution images were taken for each biometric attribute from 190 people for the study resulting in 3420 images. Preprocessing techniques such as cropping, resizing, grayscale conversion, and histogram equalization were applied to the images for uniformity. The Local Binary Patterns (LBP) technique was used for feature extraction and the resulting features were combined using the weighted average feature fusion technique. The Standard SOFM classifier was optimized using the CSO algorithm for optimal feature selection by modifying weight values. 30% of the images in the dataset were used for testing and 70% of the images in the dataset were used to train the CSO-SOFM classifier. The formulated CSO-SOFM classifier was implemented using Matlab 2016a and evaluated using metrics such as specificity, sensitivity, false positive rate, and recognition accuracy. The CSO-SOFM system obtained 98.83% accuracy, 98.83% sensitivity, 98.82% specificity, and 112.14 seconds processing time at an ideal threshold of 0.80. The findings indicate that the optimized CSO-SOFM algorithm used in this study outperformed the conventional SOFM algorithm. The SOFM algorithm was optimized by the CSO algorithm resulting in lower false positives and processing time. The approach utilized in this study provides a veritable means of enhancing access control systems and mitigating security breaches by malicious individuals in the digital space.]]>
