An automatic cryptanalysis of Arabic transposition ciphers using compression

Abstract

This paper introduces a compression-based method adapted for the automatic cryptanalysis of Arabic transposition ciphers. More specifically, this paper presents how a Prediction by Partial Matching ('PPM') compression scheme, a method that shows a high level of performance when applied to the different natural language processing tasks, can also be used for the automatic decryption of transposition ciphers for the Arabic language. Another well known compression scheme, Gzip, is also investigated in this paper with less efficient performance demonstrated by this method. In order to achieve readability, two further compression based approaches for space insertion are evaluated as well in this paper. The results of our experiments with 125 Arabic cryptograms of different lengths show that 97% of the cryptograms are successfully decrypted without any errors using PPM compression models. As well in a post-processing step, we can effectively segment the output that is produced by the automatic insertion of spaces resulting with only a few errors overall. As far as we know, this is the first work to demonstrate an effective automatic cryptanalysis for transposition ciphers in Arabic.