Name: Uddin, Abu Hamed Mohammad Misbah
Title: Privacy Preserving Collaborative Anomaly detection Using Secure Multi-party Computation
Abstract: Increasing cyber attack is a major problem to current Internet world. Collaborative intrusion detection system can help mitigating the problem to some extent. A mechanism to design such a system is aggregating attack traffic from victim organizations and applying anomaly detection system on the aggregated data. To protect privacy of the users, the organizations should aggregate in a highly secure environment. Secure multi-party computation may be applied to such a task, but the general consensus is that the computation and communication overhead of such protocols makes them impractical for aggregation of large datasets.
In our work, we present a novel way to aggregate attack traffic in privacy preserving manner using primitives of secure multiparty computation. Specifically, we have devised a protocol independent algorithm that computes fast and secure set union and intersection. We implemented our algorithm in Sharemind, a fast privacy preserving virtual computer and support our claims by experimental results.