Cold Boot Attack Methods for the Discrete Logarithm Problem

Abstract

This paper proposes improvements to existing cold boot attack algorithms which greatly reduce the number of correlate tests required at the expense of some memory in discrete logarithm based cryptosystems. In practical key recovery settings, the excess memory incurred is shown to be insignificant when the variable parameter in the algorithms is optimized. The results show that improvements of up to 2.2 times for wNAF encodings and up to 5.5 times in PolarSSL encodings can be achieved based on the number of correlate tests when such optimal values of the variable parameter are applied.