Fidelity Leakages: Applying Membership Inference Attacks to Preference Data

Danhier, Pierre;Massart, Clément;Standaert, François-Xavier
(2020) 39th IEEE Conference on Computer Communications (INFOCOM 2020) — Location: Toronto (Canada) (6.July.2020)

Files

FidelityLeakages-ApplyingMembershipInferenceAttackstoPreferenceData.pdf
  • Open Access
  • Adobe PDF
  • 318.85 KB
Download

Details

Authors
Abstract
We apply membership inference attacks in the context of preference data exploited by recommendation systems and show that they can lead to “fidelity leakages”. These leakages allow one service provider to determine whether or not its users are faithful. We first provide experimental results based on real-world data made available by Spotify that confirm the feasibility of such attacks and allow us to put forward their influencing parameters. We then discuss the challenges for interpreting and mitigating fidelity leakages.
Affiliations

Citations

  • APA
  • Chicago
  • FWB

Danhier, P., Massart, C., & Standaert, F.-X. (2020). Fidelity Leakages: Applying Membership Inference Attacks to Preference Data. Proceedings of INFOCOM 2020. 39th IEEE Conference on Computer Communications (INFOCOM 2020), Toronto (Canada). https://doi.org/10.1109/INFOCOMWKSHPS50562.2020.9163032