New Results on Cryptanalysis of Stream Ciphers

Stream ciphers are cryptographic primitives that ensure the confidentiality of communications. In this thesis, we study several attacks on stream ciphers. For practical applications, the candidates of stream ciphers of NESSIE and eSTREAM projects are scrutinized. Firstly, the algebraic attacks on SOBER-t32 and SOBER-t16 stream ciphers are performed under the assumption that the stuttering phases are not considered. In particular, our attack demonstrates that the fast algebraic attack can improve the efficiency of the algebraic attack significantly by reducing the degree of nonlinear algebraic equations with annihilators. Next, a linear distinguishing attack is presented against SOBER-128 which is a strengthened version of SOBER-t32. We observe that there is a highly biased approximation in the filter function which can be combined with the linearity of the shift register. In addition, we show that it is possible to improve the complexity of the attack by using a quadratic equation. Following this, a new type of linear distinguishing attack is applied to the NLS stream cipher. This attack demonstrates that a linear attack can be extended to stream ciphers based on a nonlinear shift register and a nonlinear filter function. The attack on NLS is extended to NLSv2 which is a tweaked version of NLS. In this attack, we show that certain combinations of multiple modular additions retain highly biased linear approximations. Finally, a distinguishing attack on the Dragon stream cipher is presented. Though this attack is far away from being practical, our attack shows that basic components of Dragon are not correlation immune. Statement of Candidate This thesis is submitted in fulfillment of the requirements of the Doctor of Philosophy at Macquarie University and has not been submitted for a higher degree to any other university or institution. This thesis represents my original work and contributions.