Design and realization of a new neural block cipher
Hassan Noura, Abed Ellatif Samhat, Youssef Harkouss, Tara Ali‐Yahiya · 2015
In this paper, we propose a new neural dynamic block cipher based on a combination between Artificial Neural Network (ANN) and an efficient nonlinear function. A dynamic construction method of synaptic weight matrices is achieved by updating the weight matrices after each validate time and the problem of reversibility is resolved. The main advantage of using the proposed structure is that it can inherent the parallel process of neural network and consequently can ensure lower computation complexity and lower energy consumption. Furthermore, theoretical and experimental results showed that the proposed cipher is characterized by a sufficient level of security and a lower latency when compared to the AES algorithm which is suitable for real time applications.