Multilayer Perceptron of Occlusion and Pose‐Sensitive Ear Attributes for Social Engineering Attack Mitigation

Abstract

Facial biometrics have emerged as a promising approach for user authentication due to their non-intrusive nature and wide availability of facial imagery. This chapter thereby focuses on enhancing user authentication in facial biometrics using multilayer perceptron for digital forensics. The study addresses the challenges posed by occlusion and pose variations, which can significantly impact the accuracy of authentication systems. By leveraging an inclusive dataset with sensitivity to identified threats in literature, the effectiveness of the Perceptron in handling these variations is investigated. One-hot encoding is employed to represent categorical data feature attributes, just as parameter optimization is implemented to maximize the model's performance. The model evaluation results demonstrate the Perceptron's effectiveness in handling occlusion and pose variations, highlighting its potential for real-world applications in digital forensics. The study emphasizes the importance of one-hot encoding and parameter optimization in improving the performance of perceptron models.