A Block Cipher Design Using Recurrent Neural Networks
As security has become a necessary component for business applications in many areas, research of new cryptography technology is desirable, especially the highly secure and efficient data encryption technique. A new block cipher designed based on recurrent neural networks is proposed for first time in the project. Recurrent neural networks have dynamics characteristics and can express functions of time. By introducing recurrent neural networks to cryptography, the proposed block cipher releases the constraint on the length of secret key. The inherited high by parallel processing capability of neural networks can also improve the encryption performance greatly. The recurrent neural networks make the block cipher strong to resist different cryptanalysis attacks and to provide data integrity and authentication service at the same time. The design of the proposed block cipher is presented and analyzed in detail. Simulation results provide illustrations. The proposed block cipher is flexible to be implemented either in software or in hardware for efficient data encryption purpose.