A Multi-party Private Set Union Protocol Against Malicious Adversary

Our protocol, grounded on the principles of SPDZ technology and the Bloom Filter, aims to efficiently compute the union of datasets from multiple participants without exposing individual participants’ data. Initially, we employ the SPDZ and the Bloom Filter to realize a secure and efficient protocol for the union computation of private sets from multiple participants. Finally, we establish the upper bounds for differential privacy budgets. We implement experiments in both LAN and WAN environments, with 10 participants and 5 CPs, on datasets of varying sizes. The results indicate that despite the huge overhead during the offline phase, this can be efficiently addressed offline, deeming our protocol’s efficiency acceptable. Furthermore, by conducting experiments on a fixed dataset within a LAN environment and varying the number of participants, we observed that the overhead cost linearly increases with the number of participants.