Convolutional Neural Network-Based Data Encryption Model for Multimedia Files Using Advanced Encryption Standard Algorithm

Abstract

With the rapid advancement of multimedia technology, securing sensitive multimedia content has become an increasingly critical challenge in the digital era. This paper presents a novel approach to multimedia encryption by developing a Convolutional Neural Network (CNN)-based data encryption system that leverages the Advanced Encryption Standard (AES) algorithm. The hybrid model addresses the growing need for robust security mechanisms to protect multimedia files against unauthorized access and cyber threats. The proposed model employs a CNN autoencoder architecture to extract meaningful features from multimedia files, which are then encrypted using the AES algorithm. Extensive performance evaluation using standard test images demonstrates that our hybrid CNN-AES model achieves an accuracy of 86% at 2000 epochs, with a throughput of 4,128,251 images per second and a latency of 3.5 seconds at 1000 epochs. The results indicate that the proposed model offers enhanced security, effective handling of data heterogeneity, and flexibility while maintaining satisfactory performance overhead for various multimedia files.