Biometric-Based Encryption System to Enhance Cloud Data Security through the Integration of Facial Recognition and Homomorphic Encryption

Abstract

The rapid growth of cloud computing has introduced significant benefits in terms of data storage and processing but it has also increased the risks of unauthorized access and data breaches. Therefore, this study presents a biometric-based encryption system which is designed to enhance cloud data security through the integration of facial recognition and homomorphic encryption. The proposed system employs an Autoencoder (AE) for feature extraction, Convolutional Neural Network (CNN) for facial recognition, and the Brakerski-Gentry-Vaikuntanathan (BGV) algorithm for secure data encryption and decryption. The adopted AE is used to efficiently compresses facial features into latent vectors used both for recognition and as encryption keys. Furthermore, the experimental evaluation of the techniques adopted using both primary facial datasets and the LFW dataset demonstrated that the AE achieved a training accuracy of 99.84% and validation accuracy of 98.59%, while the CNN attained a training accuracy of 97.05% and validation accuracy of 95.04%. Additionally, the result of the BGV encryption process recorded an average encryption time of 0.023 seconds and decryption time of 0.019 seconds, indicating minimal computational overhead. Results confirm that the integration of biometric encryption enhances both data confidentiality and authentication reliability in cloud environments. This system provides a robust and efficient framework for securing sensitive data in modern cloud infrastructures, ensuring privacy, integrity, and accessibility for authorized users.