QRES: Biologically-Inspired Secure Federated Learning for Edge IoT Devices

QRES is a deterministic, edge-native neural compression and federated learning system for time-series data. By treating compression as shared prediction and transmitting only residuals, QRES achieves high compression ratios while providing strong privacy guarantees through differential privacy, secure aggregation, and zero-knowledge proofs. The system is implemented in Rust with bit-perfect fixed-point determinism and targets constrained IoT environments.