Cryptanalyzing Merkle-Hellman Public Key Cryptosystem with Artificial Neural Networks

Analyzing any cryptographic algorithm is in general a herculean task but smart use of Artificial Intelligence(A.I.) can deliver some positive results. Various attempts regarding cryptanalysis till now have been focused on either exploiting a mathematical loop hole in the algorithm or any way out in the implementation of the algorithm. Work has been done in the field of cryptanalysis using Neural Networks and other pattern recognition techniques to identify the type of algorithm used. In this paper, we attempt to cryptanalyze the Merkle-Hellman Knapsack Public Key Cryptosystem using Shallow and Deep-multi-layered Artificial Neural Network (ANN) to know how good it is in replicating the working of the algorithm. Here we will try to investigate a third front where an ANN tries deconstruct the cryptosystem into its fundamental parts in its own way and then replicate it by learning from any relationship between cipher texts and the message or any other useful patterns it can find. Supervised learning has been used in order to check the accuracy of finding patterns and the level of classification the proposed model can perform.