An Optimization Enabled Deep Learning Based Multimodal Person Authentication System

Abstract

The practice of automatically recognizing the correct person using computational methods based on features maintained in computer systems is known as person authentication. Security, robustness, privacy, and non-forgery are the critical aspects of any person authentication system. Traditional biometric-based systems are dependent on the use of a single modality, which may be lacking in the ability to provide high security. These systems are vulnerable to noise and can be readily exploited. An optimization enabled deep learning based multimodal person authentication system is presented to solve these disadvantages. Here, a combination of brainwave signals and fingerprint images are utilized for providing improved security. A Deep Maxout Network (DMN) is utilized for performing person authentication on both the modalities and the output obtained is fused together using cosine similarity to attain the final result. The African vultures-Aquila Optimization (AVAO) algorithm is a unique optimization algorithm for updating the DMN weights. To construct the algorithm, the African Vulture Optimization Algorithm (AVOA) techniques are updated according to the extended exploration capabilities of the Aquila Optimizer (AO). The presented multimodal person authentication system achieves an accuracy of 0.926, sensitivity of 0.940, specificity of 0.928, and F1-score of 0.921, demonstrating exceptional performance. The experimental study also indicates the performance evaluation comparison of AVAO with the prevailing techniques such as Multi-task EEG-based Authentication, Multi model-based fusion, Multi-biometric system, and Visual secret sharing and super-resolution model on the basis of various metrics