Google Enhances Data Privacy with Confidential Federated Analytics - InfoQ

Google has announced Confidential Federated Analytics (CFA), a new technique that enhances privacy while increasing transparency in data processing. CFA builds on federated analytics by using Trusted Execution Environments (TEEs) to ensure only predefined computations are performed on user data without exposing raw information to servers or engineers. The system makes all privacy-relevant server-side software publicly inspectable for external verification. Google has already deployed CFA in Gboard to improve new word detection across 900+ languages, processing 3,600 missing Indonesian words in just two days compared to weeks with previous methods. CFA operates through a structured process including local data encryption, access policy enforcement, TEE execution, differential privacy algorithms, and external verification through a public transparency ledger. Google plans to extend this approach to broader federated learning tasks with integration into Android Private Compute Core.