Efficient Privacy Preserving Cross-Datasets Collaborative Outlier Detection

Outlier detection is one of the most important data analytics tasks and is used in numerous applications and domains. It is the identification of rare items, events or observations which raise suspicions by differing significantly from the majority of the data. The accuracy of the outlier detection depends on sufficient data. However, the underlying data is distributed across different organizations. If outlier detection is done locally, the results obtained are not as accurate as when outlier detection is done collaboratively over the combined data. Unfortunately, competitive advantage, privacy concerns and regulations, and issues surrounding data sovereignty and jurisdiction prevent many organizations from openly sharing their data. In this paper, we address precisely this issue. We present new and efficient protocols for privacy preserving outlier detection to find outliers from arbitrarily partitioned categorical data. Our protocols fall in the two-server model where data owners distribute their private data among two non-colluding servers who detects on the joint data using secure two-party computation (2PC). Our method is based on Local Distance-based Outlier Factor (LDOF) using the relative location of an object to its neighbours to determine the degree to which the object deviates from its neighbourhood. We provide the privacy guarantee by using secure multiparty computation techniques. We implement our system in C++ on real data. Our experiments validate that our protocols are both effective and efficient.