Attacking the Filter Generator by Finding Zero Inputs of the Filtering Function

The filter generator is an important building block in many stream ciphers. We present here an attack that recovers the initial state of the hidden LFSR by detecting the positions where the inputs of the filtering function are equal to zero. This attack requires the precomputation of low weight multiples of the LFSR generating polynomial. By a careful analysis, we show that the attack complexity is among the best known and work for almost all cryptographic filtering functions.