Machine Learning Approach for Analysing Encrypted Data

Machine Learning has been explored in many fields for solving problems ranging from medicine to finance, including Information Security. At the same time, in the area of cryptography, it is always a challenge to decipher the cryptic text without the knowledge of the encryption key. The challenge becomes even harder in absence of knowledge about the encryption algorithm. In this paper, we propose a methodology to identify the encryption algorithm being used to produce the cryptic text. We apply machine learning approach, to be specific Expectation-Maximization based clustering to identify the encryption algorithm. We conducted several experiments to substantiate our hypothesis and are effectively able to identify the encryption algorithm with over 90 percent accuracy. Our approach finds a useful place in information retrieval system when no meta-knowledge is being extended for crypt-analysis.