Neural Network Research And Application Of Block Ciphers

Along with the development of information science, information security attracts attention of international society. As the nuclear technology of information security, cryptology plays an important role in secret communication. As both are disciplines researching information processing, neural network (NN for short) will be applied in cryptology widely.In this dissertation, based on the relative theory of NN and cryptology, applications of NN in block cipher are mainly discussed at two aspects including cryptanalytics and cryptography. The main parts of the dissertation are concluded as follows:Firstly, based on design principles of block cipher and features of multi-layer feed-forward NN, the mathematical model of a block cipher is structured and realized. Simulations indicate that the block cipher is feasible. And through comparison with DES and analyses of security, the block cipher presents high degree of security, confusion and diffusion, and is easy to be realized.Secondly, different from traditional cryptanalytic approaches against public-key cryptosuystems, based on NN, an approach which destination is not decryption key is proposed. Moreover a cryptanalytic machine against RSA is structured utilizing two-layer BP NN. Simulations indicate that the eryptanalytic machine is effective and the eryptanalytic approach is feasible.Lastly, during the above process, experience and principles of designing NN oriented towards actual applications are improved and summarized.