Seminars INGI - June 5 at 1:00-2:00 p.m

Anomaly detection in CPS and my background in IoT data analysis by Mohsen Shirali ICTEAM, UCLouvain

Anomaly Detection (AD) is the process of identifying patterns in data that significantly deviate from expected behaviour. During my PhD, I focused on the Internet of Things (IoT) in healthcare applications, emphasizing systems architecture, privacy, and data analytics. I developed smart sensor systems for modelling human behaviour and assessing the health of the elderly based on physical activity analysis. Additionally, I used Process Mining and proposed techniques to detect gradual changes and anomalies in daily routines, which can indicate potential health risks.
Building on this foundation, my current research extends anomaly detection techniques to Cyber-Physical Systems (CPS). These systems integrate physical and digital components and require advanced monitoring to ensure safe and efficient operation. Anomaly detection allows for the identification of failures and unexpected behaviours. My work explores real-time anomaly detection methods, including time-series analysis, change point detection, and log analysis, as well as the application of Digital Twins to simulate and predict system behaviour. In this presentation, I will share my research background and discuss the challenges of applying anomaly detection to CPS use cases, such as vehicular networks. I will show how we are going to use the observed high redundancy in the metrics exported by vehicles to detect anomalies based on correlations.

Where : Shannon room - Maxwell, a.105,  Place du Levant 3 - 1348 Louvain-la-Neuve 

Pay attention : sandwiches will be provided. Please fill in the form before day D at 09:00 to reserve a sandwich