Adversarial Agent Collaboration for C to Rust Translation
This addresses memory safety vulnerabilities in legacy C software for developers and security practitioners, representing a novel advancement in automated translation at scale.
The paper tackles the problem of translating C to Rust for memory safety by introducing ACToR, an adversarial LLM agent-based system that successfully translates 63 real-world command-line utilities with an average size of 485 lines of code, achieving over 90% test pass rate and improving correctness by up to 18.9% compared to baselines.
Translating C to memory-safe languages, like Rust, prevents critical memory safety vulnerabilities that are prevalent in legacy C software. Existing approaches for C to safe Rust translation, including LLM-assisted ones, do not generalize on larger (> 500 LoC) C codebases because they depend on complex program analyses that frequently break. In this work, we present ACToR (Adversarial C To Rust translator), a simple LLM agent-based approach. Inspired by GANs, ACToR pits a generator agent against a discriminator agent, which collaborate to iteratively generate a Rust translation. On each iteration, the translator agent synthesizes and refines a Rust translation to pass an existing suite of tests, and then the discriminator agent finds new failing tests. We demonstrate that ACToR translates all of the 63 real-world command line utilities considered in our benchmarks, which have an average size of 485 lines of code, and it achieves over 90% test pass rate with zero human intervention. To our knowledge, it is the first such system that reliably translates C programs of this scale. Furthermore, ACToR improves translation correctness by up to 18.9% compared to baseline, non-adversarial approaches.