Normal-Only Anomaly Detection in Environmental Sensors in CPS: A Comprehensive Review

Anomaly detection is crucial for maintaining the reliability and security of environmental sensors in Cyber-Physical Systems (CPS). This paper presents a comprehensive review that examines the field of anomaly detection using only normal data (normal-only anomaly detection) within environmental sensor networks of CPS, covering literature from the past ten years. The study explores the methods, challenges, and progress made in this area offering exploration of methodologies, challenges, and advancements. With a focus on identifying deviations from normal behavior rather than categorizing specific anomalies, this review covers a wide range of anomaly detection techniques tailored to environmental sensor data. Furthermore, we offer perspectives on particular application fields of CPS and explore how anomaly detection techniques that depend exclusively on data from normal operations are implemented. We analyze the intricacies of applying machine learning, statistical, and data-driven approaches to unveil anomalies, while considering the unique characteristics and constraints of CPS environments. The review also delves into the challenges posed by noisy and dynamic sensor data, scalability issues, and the trade-offs between detection accuracy and resource efficiency. By synthesizing insights from diverse research endeavors, we provide a comprehensive understanding of the landscape, emerging directions, and potential applications of normal-only anomaly detection in environmental sensors within CPS. The study stands as an essential guide for academics, professionals, and interested parties aiming to bolster the resilience and safeguarding of CPS equipped with sensors.