Identification of Encryption Method for Block Ciphers using Machine Learning Methods

Identifying the encryption algorithm from only cipher text is a crucial task in cryptography and cryptanalysis. While some research has been done in this area using statistical methods for block ciphers, this paper proposes a novel approach using machine learning models for identifying encryption algorithms. The proposed work uses a deep neural network, random forest, and logistic multiclass regression methods and extracts features from cipher text to classify the encryption algorithm used. The model considers ECB, CBC, CFB, and CTR modes of block ciphers for the encryption of cipher text. The effectiveness of the model has been tested for AES and DES encryption algorithms. The experiments demonstrate that the proposed machine learning models significantly outperform existing statistical methods, which are limited in identifying encryption algorithms for modes other than ECB. The proposed models achieve a high success rate of more than 65 percent each in identifying encryption algorithms for various modes of block ciphers ECB, CBC, CFB, and CTR. The results of this proposed work provide a new direction for identifying encryption algorithms, particularly for modes of block ciphers other than ECB, and emphasize the prospect of cryptanalysis using machine learning methods. The proposed methods can effectively assist cryptanalysts in key recovery, improving existing results in the field. The research also underscores the importance of developing new techniques for securing communications in the digital age.