More Glimpses of the RC4 Internal State Array

Abstract

We present three categories of new non-randomness results on RC4 here. First, using Jenkins’ Glimpse result (1996), we show that for every round, the key-stream value comes from a distant permutation array (S-Box) byte position with a probability higher than the fair chance. Moreover, we show how the Jenkins’ Glimpse gets modified when the two consecutive output keystream bytes are equal to a certain value. Second, we show that corresponding to every Fluhrer-McGrew bias (2000), the key-stream value for each of the two rounds comes from certain array positions with a higher (or a lower) probability than the uniform one, depending upon whether the double-byte is positively or negatively biased. Also, there are four “lag-one digraph” (or alternate pair double-byte) biases that have been summarized in a recent paper. We show that for each of these cases, there are preferred positions of permutation array byte corresponding to the key-stream value in that round. Third, we show that in each of the Fluhrer-McGrew or lag-one cases, the Jenkins’ correlation results must be refined. Surprisingly, in one particular configuration, the Glimpse correlation value becomes almost $\frac{4}{N}$ instead of Jenkins’ $\frac{2}{N}$.