Source: CARDIS 2022
Abstract
Recent works on key rank estimation methods claim that algorithmic key rank estimation is too slow, and suggest two new ideas: replacing repeat attacks with simulated attacks (PS-TH-GE rank estimation), and a shortcut rank estimation method that works directly on distinguishing vector distributions (GEEA). We take these ideas and provide a comprehensive comparison between them and a performant implementation of a classical, algorithmic ranking approach, as well as some earlier work on estimating distinguisher distributions. Our results show, in contrast to the recent work, that the algorithmic ranking approach outperforms GEEA, and that simulation based ranks are unreliable.
