SECURE MEDICAL IMAGE ENCRYPTION IN IOT USING RNN-BASED KEY GENERATION WITH AES

Abstract

    Secure image transmission is a crucial requirement in Internet of Things (IoT) applications, particularly in the medical field, where privacy and confidentiality are of paramount importance. This paper proposed an image encryption scheme that integrated Recurrent Neural Networks (RNNs) for generating multiple cryptographic keys with the Advanced Encryption Standard (AES) to provide strong image protection. The use of RNN-based key generation enhanced the unpredictability and diversity of the keys, thereby reducing the risk of brute-force and statistical attacks. Experimental results were conducted on medical images to evaluate the scheme's performance using several security metrics. The encrypted images achieved near-ideal entropy values, approaching 8, indicating excellent randomness. PSNR values below 10 dB and high MSE confirmed strong visual distortion in the encrypted images, while the decrypted images exhibited lossless reconstruction with infinite PSNR and zero MSE. Histogram analysis showed uniform pixel intensity distribution, and correlation coefficients were close to zero, demonstrating resilience against statistical and differential attacks. the proposed RNN with AES scheme as a secure and efficient solution for IoT image encryption.