Cryptanalysis and Security Evaluation Using Artificial Neural Networks

This chapter introduces a novel methodology for determining mutual information between the secret key and cypher message. The methodology is based on using artificial intelligence, specifically neural networks, to determine correlations that can be found in the weak and medium encryption algorithms. Furthermore, this chapter presents a relevant survey in the field of research with a similar purpose, a detailed description of the developing environment, a theoretical overview of the matter, as well as a detailed methodology description with additional results. The final results show the performance of the proposed methodology, as well as the results of applying it to today’s most-secured algorithms. Based on the obtained results, this chapter proves the efficiency of the proposed methodology. Besides having Advanced Encryption Standard (AES) and Data Encryption Standard (DES) as reference algorithms, this methodology could be applied to real-world algorithms. Absolute proof of this methodology represents the simple crypt algorithm that scored over 99.9% accuracy in most dataset cases. In the end, results on AES do not show any significant accuracy of the model, but when DES comes in a case, the model scores about 70% accuracy. This result proves the real-world applications of this methodology.