Proposed neural SAE-based medical image cryptography framework using deep extracted features for smart IoT healthcare applications

Abstract

Image cryptography based on chaos algorithms is widely employed in modern security systems in telemedicine Internet of Things (IoT) applications. One of the main drawbacks of the state-of-the-art chaos encryption algorithms is that they are not sufficiently secure for image communications. When an image is only encrypted using traditional chaotic algorithms, it may be easily vulnerable to attackers. This paper presents a medical image cryptosystem, which uses a Stacked Auto-Encoder (SAE) network to produce two sets of chaotic random matrices. The first set is utilized to generate a complete shuffling matrix that changes the pixel locations in the digital input image. The second set generates an independent series of sequences employed to eradicate the correlation between the permuted encrypted medical image and the original image. The proposed cryptosystem is robust due to the benefits of parallel SAE computations, which significantly reduce the runtime and complexity. Moreover, the hybrid implementation of confusing and shuffling processes improves the ciphering performance. A comparative study between the proposed medical image cryptosystem and other related works is presented. The simulation tests reveal the security, efficiency, and immunity of the proposed cryptosystem for various forms of attacks compared to the traditional cryptosystems. The obtained results reveal that the suggested framework can be valuable and appropriate for medical sector services. It can be recommended for real-time healthcare cloud and IoT applications because of its superior accomplishments in terms of robustness, security, and computational complexity.