Reliable human authentication using AI-based multibiometric image sensor fusion: Assessment of performance in information security

Abstract

Reliable human authentication is essential for assuring the availability, confidentiality, and integrity of sensitive data and resources in the field of information security. Due to the vulnerabilities of conventional authentication techniques like passwords and PINs, interest in biometric-based authentication systems has developed. To get over this, we suggest an efficient voice and iris recognition-based multimodal biometric verification strategy for human authenticating tools. We first collect the voice and iris datasets from 150 people. Then, we used a median filter and high pass filter to preprocess the voice and iris data. The features of voice identification are extracted using the shifted delta cepstral coefficient (SDCC) and the Mel frequency discrete wavelet coefficient (MFCC), and these two coefficients are evaluated. The outcomes of the extraction of iris identification features using Local binary pattern (LBP) and Speeded robust features (SURF) are evaluated. The classifier Fine Tuned Cuckoo Search Optimized Convolutional Neural Network (FCSO-CNN) performs voice and iris recognition modalities. Voice and iris biometric systems may be combined into a single multimodal biometric system by fusing their respective feature sets and scoring algorithms. The results of the computer simulation demonstrate that for speech recognition, using the SDC and MFC coefficients produces better results, while for iris recognition, using the LBP and FCSO-CNN experiment produces superior outcomes. Additionally, the scores fusion works superior to other scenarios in the proposed multimodal biometrics system. The suggested system provides a dependable and strong solution for human identification, assisting in the improvement of security measures across a variety of industries, including financial institutions, governmental agencies, and the defense of key infrastructure.