scale-snu/cheddar-fhe

Cheddar-FHE occupies a very specific niche: high-performance FHE for Apple's ecosystem. While most FHE research focuses on C++ (Microsoft SEAL, OpenFHE) or Rust (Zama's Concrete), Cheddar targets Swift and Metal, making it uniquely suited for on-device privacy-preserving applications on iOS and macOS. Its defensibility (5) stems from the extreme technical difficulty of implementing FHE primitives like the Number Theoretic Transform (NTT) and modular reduction efficiently on GPU architectures, especially Metal. However, it lacks the massive community backing of OpenFHE or the venture funding of Zama. The primary threat is platform domination: Apple is the most likely entity to sherlock this functionality. If Apple decides FHE is a core component of their 'Private Cloud Compute' or on-device 'Apple Intelligence' strategy, they will likely release a proprietary, highly-optimized FHE framework within the Accelerate or CryptoKit libraries, rendering third-party Swift libraries obsolete. Quantitative signals (78 stars, university-led) suggest this is currently a high-quality research artifact with some developer interest but hasn't yet reached the 'infrastructure-grade' status required to resist platform absorption.