| Abstract: | Linear cryptanalysis (LC) is an important codebreaking method that became popular in the 1990s and has roots in the earlier research of Shamir in the 1980s. In this article we show evidence that linear cryptanalysis is even older. According to documents from the former East Germany cipher authority ZCO, the systematic study of linear characteristics for nonlinear Boolean functions was routinely performed in the 1970s. At the same time East German cryptologists produced an excessively complex set of requirements known as KT1, which requirements were in particular satisfied by known historical used in the 1980s. An interesting line of inquiry, then, is to see if KT1 keys offer some level of protection against linear cryptanalysis. In this article we demonstrate that, strangely, this is not really the case. This is demonstrated by constructing specific counterexamples of pathologically weak keys that satisfy all the requirements of KT1. However, because we use T-310 in a stream cipher mode that uses only a tiny part of the internal state for actual encryption, it remains unclear whether this type of weak key could lead to key recovery attacks on T-310. |
