Some Results on Distinguishing Attacks on Stream Ciphers

University dissertation from Electro and information technology

Abstract: Stream ciphers are cryptographic primitives that are used to ensure the privacy of a message that is sent over a digital communication channel. In this thesis we will present new cryptanalytic results for several stream ciphers. The thesis provides a general introduction to cryptology, explains the basic concepts, gives an overview of various cryptographic primitives and discusses a number of different attack models. The first new attack given is a linear correlation attack in the form of a distinguishing attack. In this attack a specific class of weak feedback polynomials for LFSRs is identified. If the feedback polynomial is of a particular form the attack will be efficient. Two new distinguishing attacks are given on classical stream cipher constructions, namely the filter generator and the irregularly clocked filter generator. It is also demonstrated how these attacks can be applied to modern constructions. A key recovery attack is described for LILI-128 and a distinguishing attack for LILI-II is given. The European network of excellence, called eSTREAM, is an effort to find new efficient and secure stream ciphers. We analyze a number of the eSTREAM candidates. Firstly, distinguishing attacks are described for the candidate Dragon and a family of candidates called Pomaranch. Secondly, we describe resynchronization attacks on eSTREAM candidates. A general square root resynchronization attack which can be used to recover parts of a message is given. The attack is demonstrated on the candidates LEX and Pomaranch. A chosen IV distinguishing attack is then presented which can be used to evaluate the initialization procedure of stream ciphers. The technique is demonstrated on four candidates: Grain, Trivium, Decim and LEX.