Cryptanalysis of stream cipher using density evolution

In this paper, we investigate the cryptanalysis of stream ciphers, and evaluate the enhancement to security that can be gained when the ciphertext is error prone by analyzing specific attack algorithm. The stream ciphers that we investigate here have a keystream generator that is based on linear-feedback shift registers (LFSRs). In particular, we characterize the security of these ciphers when the ciphertext obtained by an attacker is error prone. We show that a class of correlation attack can still be implemented with essentially no changes to the attack algorithms in this setting, although with limited effectiveness. The degeneration of the potency of this attacks is given as a function of the error rate in the ciphertext, and is characterized using computational security.