Classification of Encryption Algorithms Using the Hidden Markov Model

Abstract

Abtsrcat The “Hidden Markov Model”, HMM for short, represents a very rich class of stochastic models which is extensively used in modeling real data. In particular it has proved to be a strong and powerful tool for classification problems in some specific application areas including speech recognition and image analysis. The model involves a complex stochastic formulation of the underlying process and the implementation is numerically somewhat involved. In this article we present some results involving the application of hidden Markov models in trying to classify a cipher text in terms of the source algorithm generating the cipher when it is known that the source algorithm belongs to a finite list of algorithms. Our specific application is based on ten source algorithms, five each of the stream cipher and block cipher types. We perform a statistical analysis of each cipher text leading to the evaluation of the posterior chances of the cipher being generated by each of the competing source algorithms. Although there are some other instances of the use of the hidden Markov model in the area of cryptology there is little literature relating to the previous use of this model in the specific context of classification of encryption algorithms, which makes it hard to judge its performance or compare it against an existing standard. However on the basis of our limited study we feel that there is clear indication to suggest that further investigation of the hidden Markov model in the context of classification of encryption algorithms may be worthwhile as they appear to lead to classification results which are clearly better than random assignment.