Toward verifiably private learning from federated data
The paper presents a next-generation Federated Learning system that provides externally verifiable central Differential Privacy guarantees while improving device coverage and privacy-utility tradeoff.
The paper provides externally verifiable central Differential Privacy guarantees, which is a novel contribution compared to prior work.
Before reading this…
Applications
- →Training models for the Android Keyboard (Gboard)
To understand this paper, make sure you know these concepts first:
- Federated Learningfind papers →
- Trusted Execution Environmentsfind papers →
- Differential Privacyfind papers →
Abstract
More Like ThisFederated Learning (FL) allows devices with private data to collaborate in training a shared model. We present a next-generation FL system based on Trusted Execution Environments (TEEs) that addresses operational challenges associated with earlier systems and provides externally verifiable central Differential Privacy (DP) guarantees for the first time while offering a better privacy-utility tradeoff. In our system, devices upload data encrypted with keys managed by a TEE-hosted Key Management Service (KMS). The uploaded data is cryptographically tied to a policy limiting the set of Python programs that may later process the data in server-side TEEs. External parties may inspect public transparency logs to observe the set of workloads allowed by these policies. Our experimental results show that the new system improves device coverage and favorably shifts privacy-utility curves by enabling collected data to be integrated into the server-side workload at a schedule that optimizes DP guarantees and is unaffected by device availability. Our new system has been productionized, enabling models for the Android Keyboard (Gboard) to be trained faster and achieve better accuracy under smaller, now externally verifiable privacy budgets in comparison to models trained using the prior system.