The Filter-Combiner Model for Memoryless Synchronous Stream Ciphers

We introduce a new model – the Filter-Combiner model – for memoryless synchronous stream ciphers. The new model combines the best features of the classical models for memoryless synchronous stream ciphers – the Nonlinear-Combiner model and the Nonlinear-Filter model. In particular, we show that the Filter-Combiner model provides key length optimal resistance to correlation attacks and eliminates weaknesses of the NF model such as the the Anderson leakage and the Inversion Attacks. Further, practical length sequences extracted from the Filter-Combiner model cannot be distinguished from true random sequences based on linear complexity test. We show how to realise the Filter-Combiner model using Boolean functions and cellular automata. In the process we point out an important security advantage of sequences obtained from cellular automata over sequences obtained from LFSRs.