A systematic literature review of IoT time series anomaly detection solutions

The rapid spread of the Internet of Things (IoT) devices has prompted many people and companies to adopt the IoT paradigm, as this paradigm allows the automation of several processes related to data collection and monitoring. In this context, the sensors (or other devices) generate huge amounts of data while monitoring physical spaces and objects. Therefore, the problem of managing and analyzing these huge amounts of data has stimulated researchers and practitioners to adopt anomaly detection techniques, which are automated solutions to enable the recognition of abnormal behaviors occurring in complex systems. In particular, in IoT environments, anomaly detection very often involves the analysis of time series data and this analysis should be accomplished under specific time or resource constraints. In this systematic literature review, we focus on the IoT time series anomaly detection problem by analyzing 62 articles written from 2014 to 2021. Specifically, we explore the methods and techniques adopted by researchers to deal with the issues related to dimensionality reduction, anomaly localization, and real-time monitoring, also discussing the datasets used, and the real-case scenarios tested. For each of these topics, we highlight potential limitations and open issues that need to be addressed in future work.