Identification of Cryptographic Algorithm Based On Given Cipher Text

Identification of only the encryption algorithm from ciphertext is a challenging problem in cryptography and cryptanalysis. In this research we present a machine learning based method to classify encryption algorithms by extracting statistical features from the cipher text. It creates a dataset based on four symmetric encryption algorithms, namely AES, DES, RC2, and CAST, and on multiple encryptions modes like ECB, CBC, CFB, and CTR. The extracted features (Hamming weight distributions, ciphertext length, entropy-based metrics) are input into several classification models (Logistic Regression, Random Forest, Support Vector Machines, XG Boost, Light GBM and Cat Boost). The experimental outcomes show that XG Boost has the best classification accuracy of 71.0 percent, then Light GBM with 70.87 percent and Cat Boost with 70.73 percent. The results highlight the potential of machine learning to aid in the analysis of symmetric key ciphers and that further improvements in performance can be expected from feature engineering and the application of deep learning techniques. The study makes a contribution to automated methods of cryptanalysis and provides an understanding of the ability to investigate the presence of an encryption scheme and helps in formulating a more sophisticated cryptographic security framework.