Yiyuan Liu

Yiyuan Liu is an author affiliated with Zhejiang University. Liu is listed as a co-author of the paper “MorFuzz: Fuzzing Processor via Runtime Instruction Morphing enhanced Synchronizable Co-simulation”, included in the Proceedings of the 32nd USENIX Security Symposium held August 9–11, 2023 in Anaheim, California.

Research contribution

Liu’s documented publication in the provided evidence concerns processor verification and hardware fuzzing. The MorFuzz paper describes modern processors as difficult to verify completely and identifies challenges in prior hardware fuzzing techniques, including complex input grammar, deceptive mutation guidance, and model implementation differences.

MorFuzz paper

The paper proposes MorFuzz, described as a processor fuzzer for efficiently discovering software-triggerable hardware bugs. Its core idea is to use runtime information to generate instruction streams with valid formats and meaningful semantics. The work introduces a new input structure supporting multi-level runtime mutation primitives, an instruction morphing technique for dynamic instruction mutation, and a state synchronization technique for synchronizable co-simulation across microarchitectures.

The evaluation reported in the paper covered three open-source RISC-V processors: CVA6, Rocket, and BOOM. The authors reported discovering 17 new bugs, with 13 CVEs assigned.