Key classification attack on block ciphers

TL;DR

This paper presents a key classification attack on block ciphers with longer keys than block length, enabling key recovery by exploiting statistical properties and key classes, demonstrated on KASUMI.

Contribution

It introduces a novel key classification attack method for block ciphers with large key spaces, utilizing statistical distribution assumptions and key classes for efficient key recovery.

Findings

Successfully applied to 2-round KASUMI cipher

Achieved key recovery with complexity O(max(2^n, 2^{k-n}))

Demonstrated vulnerability due to key class structure

Abstract

In this paper, security analysis of block ciphers with key length greater than block length is proposed. When key length is significantly greater than block length and the statistical distribution of cipher system is like a uniform distribution, there are more than one key which map fixed input to fixed output. If a block cipher designed sufficiently random, it is expected that the key space can be classified into same classes. Using such classes of keys, our proposed algorithm would be able to recover the key of block cipher with complexity O(max(2^n, 2^{k-n}) where n is block length and k is key length. We applied our algorithm to 2- round KASUMI block cipher as sample block cipher by using weakness of functions that used in KASUMI.