中图分类号： TP391
文献标识码： A
DOI： 10.20044/j.csdg.2097-1788.2022.03.006
引用格式： 刘节威，王钢，颜培志，等. 基于CNN的国产商用分组密码算法识别研究[J].网络安全与数据治理，2022，41(3)：33-39.

Research on identification of domestic commercial block cipher algorithms based on CNN

Abstract： With the popularization of domestic commercial cryptographic algorithms and the advancement of commercial cryptographic application evaluation, the application compliance of commercial cryptographic algorithms has attracted much attention. This paper attempts to carry out identification research on commercial block cipher algorithms and apply them to the evaluation of commercial cipher applications. In this paper，a block cipher algorithm identification scheme based on the combination of autoencoder and Convolutional Neural Networks(CNN) is proposed. The NIST randomness test method is used to extract the features of the ciphertext, and finally the ciphertext features are trained and tested by means of CNN. Experimental results show that whether the key lengths of the cryptographic algorithms are consistent is an important factor affecting the recognition accuracy. The accuracy of SM4 and other algorithms can reach more than 80%， and compared with the existing schemes, the recognition effect has higher accuracy and stability.

Key words : randomness test；cryptographic algorithm identification；convolutional neural network；autoencoder；SM4 algorithm