Ciphertext Analysis Method Based on Skip Connection and Residual Autoencoder

This research paper proposes a ciphertext analysis method that based on skip connections and residual autoencoders. By reference the skip connection approach of the U-Net model, we improve the traditional denoising autoencoder. The denoising autoencoders have achieved good performance in the task of ciphertext deciphering. However, its original structure has certain limitations, which restrict the further improvement of performance. In order to solve this question, this paper introduces residual connections in the hidden layer, and adds skip connections between the corresponding layers of the encoder and the decoder at the same time to construct an architecture similar to U-Net. The residual module enhances the ability of information fusion and detailed feature expression, which is helpful for extracting more discriminative high-order statistical features. The skip connection strengthens the cross-layer information transmission and multi-scale feature fusion, improving the decoder's ability to retain key information and texture details in plaintext reconstruction. In addition, in this paper, taking the 2D-SCL (2D chaotic map based on the sine map, the chebyshev map and a linear function) encryption algorithm as an example, the proposed ciphertext evaluation method was experimentally verified. The experimental results show that, compared with the denoising autoencoder, this method can effectively reduce information loss and improve the similarity of deciphered plaintext. At the end, we explored the performance of the skip connections and the residual autoencoder model in the application of ciphertext analysis, and proposed some improvement suggestions.