FED-chi2: Privacy Preserving Federated Correlation Test

In this paper, we propose the first secure federated chi2-test protocol Fed-chi2. To minimize both the privacy leakage and the communication cost, we recast chi2-test to the second moment estimation problem and thus can take advantage of stable projection to encode the local information in a short vector. As such encodings can be aggregated with only summation, secure aggregation can be naturally applied to hide the individual updates. We formally prove the security guarantee of Fed-chi2 that the joint distribution is hidden in a subspace with exponential possible distributions. Our evaluation results show that Fed-chi2 achieves negligible accuracy drops with small client-side computation overhead. In several real-world case studies, the performance of Fed-chi2 is comparable to the centralized chi2-test.