Hardening Confidential Federated Compute against Side-channel Attacks
arXiv:2603.2146933.1h-index: 3Has Code
AI Analysis
This addresses security risks for federated learning systems, but it is incremental as it builds on existing DP and platform work.
The paper tackled side-channel vulnerabilities in a Confidential Federated Compute platform that could undermine differential privacy guarantees, showing how DP can mitigate two of these side-channels with one implemented in an open-source library.
In this work, we identify a set of side-channels in our Confidential Federated Compute platform that a hypothetical insider could exploit to circumvent differential privacy (DP) guarantees. We show how DP can mitigate two of the side-channels, one of which has been implemented in our open-source library.