Privacy-Preserving Collaborative Anomaly Detection for Participatory Sensing

In collaborative anomaly detection, multiple data sources submit their data to an on-line service, in order to detect anomalies with respect to the wider population. A major challenge is how to achieve reasonable detection accuracy without disclosing the actual values of the participants’ data. We propose a lightweight and scalable privacypreserving collaborative anomaly detection scheme called Random Multiparty Perturbation (RMP), which uses a combination of nonlinear and participant-specific linear perturbation. Each participant uses an individually perturbed uniformly distributed random matrix, in contrast to existing approaches that use a common random matrix. A privacy analysis is given for Bayesian Estimation and Independent Component Analysis attacks. Experimental results on real and synthetic datasets using an auto-encoder show that RMP yields comparable results to non-privacy preserving anomaly detection.