Privacy-preserving anomaly counting for time-series data in edge-assisted crowdsensing

Crowdsensing is an emerging data collection paradigm that enables data collected from a large number of Internet of Things devices to support effective decision-making. Anomaly counting as a data analysis method allows the identification of unintended behaviors to enhance decision-making capabilities. However, ensuring the sensing data privacy and increasing the willingness of data providers are significant challenges to guarantee quality decision-making. This paper proposes a flexible mechanism to provide the service of privacy-preserving anomaly counting for time-series data in edge-assisted crowdsensing. Specifically, to protect the sensing data of the data providers, a secure secret sharing protocol is designed based on additive secret sharing. Next, a privacy-preserving anomaly counting algorithm based on the windowed Gaussian anomaly detector is proposed, and multiple secure sub-protocols are employed as building blocks to guarantee the privacy of the counting result and the sensing data. Additionally, the algorithm supports flexible setting of the metric of anomaly detection by the data requester when the anomaly score of sensing data is protected. Security analysis proves that the proposed scheme protects the sensing data and the results of anomaly counting for data providers and the data requester respectively. A series of experiments based on two real datasets and smart devices demonstrate that the proposed scheme is effective and saves more than half of the computation, communication, and storage cost for data providers.