This research paper addresses the issue of maintaining privacy budget in differentially-private (DP) systems. Authors propose ElephantDP, a system designed to provide the same guarantees as a trusted curator in the global DP model in an untrusted environment. ElephantDP relies on a state continuity module and a Trusted Execution Environment (TEE) to execute DP code and update the privacy budget. The paper demonstrates that this system can prevent attacks that exploit failure to maintain a privacy budget, such as replay, rollback, and fork attacks. The system has been tested using Intel SGX as a TEE, showing only a minimal overhead compared to an insecure baseline.
Publication date: 1 Feb 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2401.17628
