Secure Transmission of Images Based on Chaos Encryption and Deep Neural Network

Abstract

In today’s world of digital media, digital images are the most frequently used method of transmitting images, and therefore, their protection is significant. As it has been seen, the technology keeps on changing and thus, though there have been improvements in encryption of images, the problem of secure transmission is not completely solved. Therefore, the process must be made continuous and refined to protect data privacy and integrity, as well as transmission and storage reliability. This work focuses on the application of cryptography together with neural networks employing Python with Keras for secure image transfer. The proposed method involves image encryption through pixel shuffling, logistic-map chaotic values, and XOR operations to produce encrypted images. The neural network layer decodes the encrypted images and measures the quality of the decrypted images dependent on PSNR, MSE, and Correlation Coefficients. Thus, by comparing these metrics for the original and reconstructed images, the project will improve security and determine the effectiveness of the encryption-decryption method. The combination of the chaotic maps and neural networks gives a strong foundation for image encryption. The chaotic values bring in randomness into the encryption process, thus increasing the level of security while the neural network helps in perfect reconstruction and image quality. This approach is advantageous as it addresses both the issues of data security and, after decryption, image quality and so it offers a midway point for secure image transmission in this technology-driven world.