An Optimized and Privacy-Preserving System Architecture for Effective Voice Authentication over Wireless Network

Abstract

The speaker authentication systems assist in determining the identity of speaker in audio through distinctive voice characteristics. Accurate speaker authentication over wireless network is becoming more challenging due to phishing assaults over the network. There have been constructed multiple kinds of speech authentication models to employ in multiple applications where voice authentication is a primary focus for user identity verification. However, explored voice authentication models have some limitations related to accuracy and phishing assaults in real-time over wireless network. In research, optimized and privacy-preserving system architecture for effective speaker authentication over a wireless network has been proposed to accurately identify the speaker voice in real-time and prevent phishing assaults over network in more accurate manner. The proposed system achieved very good performance metrics measured accuracy, precision, and recall and the F1 score of the proposed model were98.91%, 96.43%, 95.37%, and 97.99%, respectively. The measured training losses on the epoch 0, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100 were 2.4, 2.1, 1.8, 1.5, 1.2, 0.9, 0.6, 0.3, 0.3, 0.3, and 0.2, respectively. Also, the measured testing losses on the epoch of 0, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100 were 2.2, 2, 1.5, 1.4, 1.1, 0.8, 0.8, 0.7, 0.4, 0.1 and 0.1, respectively. Voice authentication over wireless networks is serious issue due to various phishing attacks and inaccuracy in voice identification. Therefore, this requires huge attention for further research in this field to develop less computationally complex speech authentication systems.