Batch Image Encryption and Compression using Chaotic Map Infused Autoencoder Network
Suchana Das, Anmol Gautam, Surmila Thokchom, Bunil Kumar Balabantaray · 2022 IEEE 9th Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) · 2022
Image encryption has been an active research area in computer science. Secure transmission of images over any transmission medium is very important to ensure security and privacy. In this pursuit, many algorithms have been proposed that use chaotic map-based encryption followed by the use of Autoencoders to compress the image in a lossless manner. We have proposed an Encoder-Decoder-based method that makes use of chaotic map infusion during the training process along with OTP to generate an encrypted compressed representation of the image. The encrypted representation is then flattened to generate a one-dimensional sequence that is transmitted over the transmission medium. The trained encoder is used to generate this encrypted representation and the trained decoder is used to regenerate the original image. The advantage of this method is twofold: the spatial information is totally removed in encrypted representation and the representation is compressed in a lossless manner. The comparative analysis is done to examine the similarity between the decrypted and plain images to ascertain the efficacy of the proposed model. On evaluation, it is found that the PSNR, MSE, and SSIM values are surpassing the other related work on most Standard Test Images. The proposed model algorithm provides High PSNR (>80dB), Low MSE (>0.00007), High SSIM (>99%) in comparison with other existing neural network-based algorithms on Standard Test Images. The proposed model achieves high-performance scores on entire datasets as well. The experimental results suggest that using deep learning-based architecture can provide fruitful results in cryptography.