Cryptographic Algorithm Identification Using Machine Learning and Massive Processing

This paper presents a study on encryption algorithms identification by means of machine learning techniques. Plain text files, written in seven different languages, were encoded by seven cryptographic algorithm under ECB mode. The resulting cryptograms were submitted to a transformation so that it was possible to create metadata files. These files provide information for six data mining algorithms in order to identify the cryptographic algorithm used for encryption. The identification performance was evaluated and the language influence at the procedure was analyzed. The overall experiment involves many cryptograms, a great quantity of metadata, a huge time consuming computation, and therefore, it was employed a high performance computer. The successful identification for each mining algorithm is greater than a probabilistic bid, and there are several scenarios where algorithm identification reaches almost full recognition.