BWhat changed
Connected encrypted client uploads to predeclared access policies, attested TEE workloads and a stateful key-management ledger; later training deployment added distributed Python execution and controlled recovery while optimizing device participation.
Scientific scope
Connected encrypted client uploads to predeclared access policies, attested TEE workloads and a stateful key-management ledger; later training deployment added distributed Python execution and controlled recovery while optimizing device participation. Bundles 2024 architecture and 2026 training-system evolution with separate author sets. Unlike classical FedAvg, encrypted training examples may leave devices. Confidentiality/attestation and privacy accounting depend on TEE hardware, trusted code, policy and state continuity; side channels and hardware vulnerabilities remain outside stated protections. Open-source inspectability is not a formal proof that every workload implements DP. Arbitrary Python is responsible for anonymizing released results. Deployment/utility claims are provider-reported, not independently replayed.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Bundles 2024 architecture and 2026 training-system evolution with separate author sets. Unlike classical FedAvg, encrypted training examples may leave devices. Confidentiality/attestation and privacy accounting depend on TEE hardware, trusted code, policy and state continuity; side channels and hardware vulnerabilities remain outside stated protections. Open-source inspectability is not a formal proof that every workload implements DP. Arbitrary Python is responsible for anonymizing released results. Deployment/utility claims are provider-reported, not independently replayed.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Confidential Federated Computations
- Toward provably private learning from federated data
- Google Research: Toward provably private learning from federated data
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction