A Utility-Optimized Mechanism for Private Data Aggregation
Hang Fu, Zhengwei Lei, Minli Zhang · 2021
With the era of big data, it becomes ubiquitous to aggregate and analyze data from millions of users. However, privacy is a critical issue in data aggregation, since data contributed by users may disclose sensitive information about individuals. Local differential privacy (LDP) is emerging as the de facto concept for private data aggregation without any reliance on trusted data aggregator or third-parties. In LDP mechanisms, each user locally perturbs his raw data employing a randomized mechanism and sends the perturbed version to the data collector. In this paper, we study the problem of privacy-preserving average aggregation for count value and propose a novel local differential private mechanism: Bit Frequency Mechanism (BFM). We provide both theoretical and experimental analyses of the proposed mechanism, specifically, the experimental results demonstrate that our mechanism can averagely reduce 50% of -error compared to existing state-of-the-art approaches.