Divergence Bounds on Key Equivocation and Error Probability in Cryptanalysis

A general method, based on the f-divergence (Csiszar) is presented to obtain divergence bounds on error probability and key equivocation. The method presented here is applicable for discrete data as well as for continuous data. As a special case of the f-divergence it is shown that the upper bound on key equivocation derived by Blom is of the Bhattacharyya type. For a pure cipher model using a discrete memoryless message source a recursive formula is derived for the error probability. A generalization of the s-unicity distance is giver, from which it is shown why the key equivocation is a poor measure of theoretical security in many cases, and why lower bounds on error probability must be considered instead of upper bounds. Finally the concept of unicity distance is generalized in terms of the error probability and is called the PeSecurity Distance.