Privacy-preserving anomaly detection in the cloud for quality assured decision-making in smart cities

Rapid urbanisation places extensive demands on city services and infrastructure that mandate innovative and sustainable solutions which increasingly involve streamlined monitoring, collection, storage and analysis of massive, heterogeneous data. Analytics services, such as anomaly detection, work to both extract knowledge and support decision-making mechanisms that enable smart functionality over such contexts. However, data privacy and data quality remain significant challenges to assuring the quality of decision-making. This paper introduces a scalable, cloud-based model to provide a privacy preserving anomaly detection service for quality assured decision-making in smart cities. Homomorphic encryption is employed to preserve data privacy during the analysis and MapReduce based distribution of tasks and parallelisation is used to overcome computational overheads associated with homomorphic encryption. Experiments demonstrate that a high level of accuracy is maintained for anomaly detection performed on encrypted data with the adopted distributed data processing approach significantly reducing associated computational overheads. (C) 2018 Elsevier Inc. All rights reserved.