SECURITY OF SENSITIVE DATA IN FACE RECOGNITION SYSTEM APPLICATIONS: A NOVEL ENCRYPTION APPROACH

Face recognition has emerged as a key biometric authentication method, widely used in security-sensitive applications like mobile banking and online transactions. However, the security and privacy of facial data remain significant concerns. This project proposes a novel security framework that integrates DeepFace, a deep learning-based face recognition library, with AES (Advanced Encryption Standard) encryption to protect facial embeddings, ensuring secure storage and transmission of sensitive biometric data. Facial embeddings, extracted from input images by DeepFace, are encrypted using AES in CBC mode. This encryption method leverages dynamic key generation, ensuring unique encryption for each communication and significantly reducing the risk of unauthorized access. The system is evaluated using a dataset of facial images, and its effectiveness is tested in terms of face recognition accuracy and security. The encryption process does not impact the system's performance, achieving a recognition accuracy of 90.1% and operating with constant time complexity (O(1)), making it suitable for real-time applications. The proposed method addresses the critical privacy and security challenges in face recognition systems, providing a reliable solution for mobile authentication, banking applications, and other sensitive environments.

Keywords: Face recognition, DeepFace, AES encryption, facial embeddings, CBC mode, biometric security, privacy, mobile authentication, real-time applications, security framework