Secure and Differentially Private Detection of Net Neutrality Violations by Means of Crowdsourced Measurements

Abstract

Evaluating Network Neutrality requires comparing the quality of service experienced by multiple users served by different Internet Service Providers. Consequently, the issue of guaranteeing privacy-friendly network measurements has recently gained increasing interest. In this paper we propose a system which gathers throughput measurements from users of various applications and Internet services and stores it in a crowdsourced database, which can be queried by the users themselves to verify if their submitted measurements are compliant with the hypothesis of a neutral network. Since the crowdsourced data may disclose sensitive information about users and their habits, thus leading to potential privacy leakages, we adopt a privacy-preserving method based on randomized sampling and suppression of small clusters. Numerical results show that the proposed solution ensures a good trade-off between usefulness of the system, in terms of precision and recall of discriminated users, and privacy, in terms of differential privacy.