Research On Grouping Encryption Algorithm Recognition Based On Machine Learning

Cryptanalysis is an important way to decipher information and ensure network information security.Identifying the encryption algorithm of the ciphertext is the premise of further cryptanalysis.Block encryption is widely used and is the mainstream encryption method at present.Therefore,the research on identification block encryption algorithm is of great significance to cryptanalysisThe existing research on the identification of block encryption algorithm is lack of identifying the plaintext scenarios of different types of plaintext and the ciphertext obtained by the encryption algorithm in different working modes.Based on the method of machine learning,this paper focuses on the identification of block encryption algorithm,and puts forward a variety of new identification schemes.The main work and innovations are as follows:(1)Ciphertext feature analysis is an important prerequisite for the study of identifying block cipher algorithms.Based on the definition of information entropy,this paper proposes a new method to construct ciphertext features.Through the character information entropy,the mean value deviation degree and the extreme value deviation degree of information entropy,12 kinds of ciphertext features are constructed,and the constructed features are used in subsequent experiments to obtain better recognition results,which verifies the effectiveness of the method proposed in this paper.(2)In order to study the influence of different plaintext scenarios on the recognition of block encryption algorithm,a recognition scheme of block encryption algorithm based on K-means + + is proposed in this paper.In this scheme,four different kinds of plaintext strings are encrypted by five block encryption algorithms,and two ciphertext features with the largest information gain are extracted as the sample feature set.Compared with the K-means,the recognition accuracy of K-means + + in each plaintext scene is improved,and the scheme has better recognition effect on AES,3DES and SM4 block encryption algorithms.(3)In order to expand the gap between ciphertext features and improve the recognition accuracy of block encryption algorithm,an improved C4.5 decision tree algorithm.Combining Delphi and C4.5 decision tree,giving different weight coefficients of ciphertext characteristics,and using the ciphertext encrypted by five grouping encryption algorithms to verify the improved decision tree.The results show that the improved C4.5 Compared with the original algorithm,the recognition accuracy of 3DES,AES,Blowfish,IDEA and SM4 algorithm is improved by 17%,16%,11%,12% and 21% respectively,which verifies the effectiveness of the scheme.(4)In order to study the influence of the working mode of encryption algorithm on the identification of 5g encryption algorithm,this paper proposes a 5g encryption algorithm identification scheme based on Adaboost to identify the encryption algorithm in different working environments and different working modes of the same algorithm.The experimental results show that the recognition accuracy of 5g encryption algorithm under the same working mode is high,up to 82.62%.The encryption algorithm recognition accuracy of ECB working mode is high under different working modes.In the recognition of working mode,the accuracy of ECB working mode recognition is high.