Graph-Based Approach to the Edit Distance Cryptanalysis of Irregularly Clocked Linear Feedback Shift Registers

This paper proposes a speed-up of a known-plaintext attack on some stream ciphers based on Linear Feedback Shift Registers (LFSRs). The algorithm consists of two basic steps: first, to guess the initial seed value of one of the LFSRs, and then to use the resulting binary sequence in order to deduce useful information about the cipher parameters. In particular, the proposed divide-and-conquer attack is based on a combination of graph-based techniques with edit distance concepts. While the original edit distance attack requires the exhaustive search over the set of all possible initial states of the involved LFSR, this work presents a new heuristic optimization that avoids the evaluation of an important number of initial states through the identifi- cation of the most promising branches of the search graph. The strongest aspects of the proposal are the facts that the obtained results from the attack are absolutely deterministic, and that many inconsistent initial states of the target LFSRs are recognized and avoided during search.