Joint Optimization–Encryption Model Based on Auto Encoder Neural Network, Variable-Length Dynamic S-Box and New Stream Encryption

Abstract

The openness of the Internet makes information transmission prone to the risk of illegal eavesdropping, and data security and user privacy protection have become one of the key research hotspots. Aiming to solve the problem that a single encryption method cannot provide comprehensive security guarantees and the traditional one-time pad encryption method consumes a large amount of secure channel resource, this paper innovatively proposes a variable-length dynamic S-box and a cross-mapping sequence-based new type of stream encryption, and an improved method of synchronizing keys during a one-time pad transmission. On this basis, this paper innovatively proposes a joint optimization–encryption model based on the auto encoder neural network, variable-length dynamic S-box and new stream encryption. Finally, taking the optical fiber transmission as an example, the simulation results show that the joint optimization–encryption model proposed in this paper can save security channel resources and effectively resist common security attacks in the process of high-speed information transmission.