Harnessing the Power of LLM to Support Binary Taint Analysis

Puzhuo Liu, Chengnian Sun, Yaowen Zheng +8

This paper proposes LATTE, the first static binary taint analysis that is powered by a large language model (LLM). LATTE is superior to the state of the art (e.g., Emtaint, Arbiter…