Implementing Computer Vision and Biometrics into User Authentication

The highly digital world that is present today requires robust protection of personal information, primarily when it is handled digitally. As cyber threats continually expand, the demand for reliable authentication systems becomes more critical. Many security systems are more secure than currently implemented since some still use basic and multi-factor authentication features. Unfortunately, these systems are tedious to navigate and would compromise user experience so that it can be more secure. This paper goes through research in the field, focusing on facial and speech recognition. Additionally, implementing current authentication systems will be evaluated and used further as benchmarking. The review intends to gather an understanding of the current state of research and real-world implementation so that a method of implementing computer vision in biometric authentication can be proposed. This paper comprehensively overviews the current state-of-the-art facial and brief speech recognition state. Training a model and evaluating its accuracy can be viable for biometric authentication. This paper first demonstrates facial recognition as Labelled Faces in the Wild (LFW) taken from Kaggle.com. The proposed result was focused on the accuracy metric. This paper shall be continued by using libraries such as Keras Tuner and Optuna to assist in selecting the optimal set of hyperparameters.