Image Cryptography Enhancement of the Arnold Cat Map Algorithm Using Long Short-Term Memory Generated Sequence

The recent massive shutdowns show that we need to guard against cyberspace as if it were a battlefield. Thus, one must consider the advancement of data encryption and decryption techniques. For enhancing the ACM algorithm, we propose in this study a dual-stream network with a generated sequence network based on Long Short-Term Memory (LSTM). The main purpose of the integration was to overcome the inherent drawbacks of predictability and non-periodic pixel reshuffling. Results show superior performance with the enhanced ACM generated via an LSTM, with higher entropy (5.67) and a lower correlation coefficient (0.006), compared with the original ACM, which has lower entropy (5.30) and a higher correlation coefficient (0.013). Additionally, the enhanced ACM exhibits a stronger, more substantial effect (130.52) than the original ACM (66.25), suggesting greater resistance to differential cryptanalysis. Therefore, the proposed enhancement of the ACM via LSTM-generated sequences improves security by increasing pixel unpredictability and randomness and by distorting image information.