|Xuhui Gong||Huazhong University of Science and Technology, P.R. China|
|Qiangsheng Hua||Huazhong University of Science and Technology, P.R. China|
|Lixiang Qian||Huazhong University of Science and Technology, P.R. China|
|Dongxiao Yu||Huazhong University of Science and Technology, P.R. China|
|Hai Jin||Huazhong University of Science and Technology, P.R. China|
Privacy-preserving data aggregation has been extensively studied in the past decades. However, most of these works target at specific aggregation functions such as additive or multiplicative aggregation functions. Meanwhile, they assume there exists a trusted authority which facilitates the keys and other information distribution. In this paper, we aim to devise a communication efficient and privacy-preserving protocol that can exactly compute arbitrary data aggregation functions without trusted authority. In our model, there exist one untrusted aggre-gator and n participants. We assume that all communication channels are insecure and are subject to eavesdropping attacks. Our protocol is designed under the semi-honest model, and it can also tolerate k (k ≤ n − 2) collusive adversaries. Our protocol achieves (n−k)-source anonymity. That is, for the source of each collected data aparting from the colluded participants, what the aggregator learns is only from one of the (n − k) non-colluded ones. Compared with recent work  that computes arbitrary aggregation functions by collecting all the participants' data using the trusted authority, our protocol increases merely by at most a factor of O((log n log log n) 2) in terms of computation time and communication cost. The key of our protocol is that we have designed algorithms that can efficiently assign unique sequence numbers to each participant without the trusted authority.