Private Rank Aggregation under Local Differential Privacy
share
In typical collective decision-making scenarios, rank aggregation aims to combine different agents' preferences over the given alternatives into an aggregated ranking that agrees the most with all the preferences. However, since the aggregation procedure relies on a data curator, the privacy within the agents' preference data could be compromised when the curator is untrusted. All existing works that guarantee differential privacy in rank aggregation assume that the data curator is trusted. In this paper, we first formulate and address the problem of locally differentially private rank aggregation, in which the agents have no trust in the data curator. We propose an effective and efficient protocol LDP-KwikSort, with the appealing property that each agent only needs to answer a small number of pairwise comparison queries from the untrusted curator with controllable noise, and the aggregated ranking could maintain an acceptable utility compared with that of the non-private protocol. Theoretical and empirical results demonstrate that the proposed solution can achieve the practical trade-off between the utility of aggregated ranking and the privacy protection of agents' pairwise preferences.
READ FULL TEXT