An analysis of the statistical self-synchronization of stream ciphers

In this paper, we examine a recently proposed mode of operation for block ciphers which we refer to as statistical cipher feedback (SCFB) mode. SCFB mode configures the block cipher as a keystream generator for use in a stream cipher such that it has the property of statistical self-synchronization, thereby allowing the stream cipher to recover from slips in the communications channel. Statistical self-synchronization involves feeding back ciphertext to the input of the keystream generator similar to the conventional cipher feedback (CFB) mode of block ciphers, except that the feedback only occurs when a special pattern is recognized in the ciphertext. In the paper, we examine the efficiency, resynchronization, and error propagation characteristics of SCFB and compare these to the conventional modes of CFB, output feedback (OFB), and counter mode. In particular, we study these characteristics of SCFB as a function of the synchronization pattern size. We conclude that, although it can take significantly longer to resynchronize, SCFB mode can be used to provide self-synchronizing implementations for stream ciphers that are much more efficient than conventional CFB mode and that have error propagation characteristics similar to CFB mode