Convolution-based Image Encryption for computer networks

In this paper, a new method for secure image transmission in computer networks is presented, utilizing convolution-based image encryption. The proposed method employs an asymmetric model, which combines adversarial training and image compression techniques to encrypt and compress the image data. The encryption process involves transforming the image into encrypted feature maps through a series of convolutional layers. These feature maps are then compressed using lossy compression techniques to minimize the data size. The decryption process reverses the encryption steps to recover the original image. An evaluation of the proposed method is carried out on standard image datasets to demonstrate its effectiveness. The decoder uses the PSNR metric to evaluate the image quality, and we achieve a PSNR score of ۴۴, indicating high fidelity to the original image. The results demonstrate that the proposed method outperforms existing image encryption and compression techniques in terms of both security and compression performance. In conclusion, the proposed method offers a promising solution for secure image transmission in computer networks.

Keywords:

Image compression , image encryption , deep learning , adversarial training , asymmetric model