Cryptanalysis of LOKI

Abstract

In [BrPiSe90] Brown, Pieprzyk and Seberry proposed a new encryption primitive, which encrypts and decrypts a 64-bit block of data using a 64-bit key. Furthermore they propose a way to build private versions of LOKI.

In this paper we show first that the keyspace of any LOKI-version is only 2⁶⁰, not 2⁶⁴ as claimed. Therefore there are 15 equivalent keys for every key, that encrypts/decrypts texts the same way. An immediate consequence is, that the proposed Single Block Hash Mode is no good. It is very easy to find collisions.

Secondly we do differential cryptanalysis on LOKI and show that n-round LOKI, n≤14 is vulnerable to this kind of attack, at least in principle. We show that we cannot find a characteristic with a probability high enough to break LOKI with 16 rounds. However one might find a private LOKI-version, that is vulnerable to a differential attack for n=16.

The main part of this paper was made at the Technical University of Eindhoven, the Netherlands, during my stay there from 1.2.91 to 29.6.91.