fabecode/GraphML-FHE

GraphML-FHE is a classic academic reference implementation associated with the 14th SPACE Conference. With only 10 stars and a 2-year period of inactivity (velocity 0.0), it serves as a 'frozen' proof-of-concept rather than a living project. The defensibility is low because the code lacks a supporting ecosystem, performance optimizations required for production FHE (which is notoriously slow), and active maintenance. From a competitive standpoint, companies like Zama (Concrete ML) and Duality Technologies are building robust, production-grade FHE compilers that can handle GNNs more efficiently than a standalone research script. While frontier labs (OpenAI/Google) are unlikely to build niche AML-specific FHE tools, the risk of displacement comes from specialized PPML (Privacy-Preserving Machine Learning) startups that are moving toward general-purpose FHE frameworks. The project's value lies entirely in the domain-specific application of GNNs to AML datasets in an encrypted domain, but as an open-source asset, it lacks the 'gravity' or 'moat' to prevent being superseded by more modern, optimized libraries like TenSEAL or Concrete.