eth-sri/dp-sniper

dp-sniper is a research artifact from the prestigious SRI lab at ETH Zurich. While technically sound at the time of publication (circa 2019), it suffers from the 'academic rot' typical of research code. With only 23 stars and zero activity in years, it functions more as a proof-of-concept for a specific paper than a maintained tool. The core technique—using an ML classifier as an adversary to find lower bounds for epsilon—is now a standard approach in DP auditing. It has been effectively superseded by more modern and actively maintained libraries such as Google's 'dp-auditorium' or the 'OpenDP' project. The defensibility is low because the implementation is easily replicated and the project lacks any community traction or integration ecosystem. Frontier labs are unlikely to compete here as this is a niche security/compliance tool rather than a core generative AI capability; however, the project is already displaced by newer academic and corporate research in the privacy-preserving ML space.