Privacy-preserving anomaly detection in cloud with lightweight homomorphic encryption

Abstract

Anomaly detection on large-scale, complex and dynamic data is an essential service that is vital to enable smart functionality in most systems. Increased reliance on cloud computing infrastructures to process such data pose critical challenges with regard to security and privacy. This paper introduces a practical framework that takes advantage of cloud resources to provide a lightweight and scalable privacy preserving anomaly detection service for sensor data. A lightweight Homomorphic Encryption scheme is used to ensure data security and privacy with any computational limitations overcome through a convenient data processing model that employs a single private server collaborating with a set of public servers within a cloud data centre. Virtual nodes implemented on public servers perform granular anomaly detection operations on encrypted data. Comprehensive experimentation demonstrates consistently high detection accuracy with less overheads in a cloud-based anomaly detection model that is both lightweight and scalable while ensuring data privacy.