Context Aware Security for Industrial Control Systems
Project Overview
Industrial Cyber-Physical Systems (ICPSs) manage critical manufacturing processes where digital controls and physical behaviour are tightly coupled. Sophisticated adversaries exploit this connection by manipulating sensor readings while controlling connected devices, evading traditional network traffic analysis. CASPER (Context Aware Security for cyber Physical Edge Resources) addresses this vulnerability by deploying a secondary, low-cost IoT sensor network that operates on separate communication protocols such as Bluetooth and ZigBee, creating an air-gapped layer of protection independent from primary control channels. The project pursues four main objectives: reviewing current ICPSs from cybersecurity perspectives, developing a reconfigurable IoT sensing infrastructure for analytics deployment, augmenting cyberattack detection through physical and behavioural monitoring, and evaluating the effectiveness of a context-aware, dynamically adaptive IoT edge network for protecting manufacturing environments.
Sensors positioned around industrial robotic arms observe temperature, vibration, light, and sound to detect anomalies. Edge analytics and state-of-the-art deep learning correlate these physical signatures with expected behaviour, enabling operators to receive early warnings before faults escalate, even when attackers spoof primary telemetry. This physical-layer monitoring complements traditional network defences by providing an independent source of ground truth.
The work has produced the CASPER platform, validated through experimental deployments. The project has contributed open datasets and toolkits for the broader research community, supporting reproducible experimentation in cyber-physical anomaly detection for industrial environments.
Team
Partners
Repositories
industrial-robotics-anomaly-detection
Real-time anomaly detection in industrial robotic arms via TinyML. Includes scripts for cloud and edge computing, datasets, and Jupyter notebooks for data analysis and model demonstrations using the Nicla Sense ME with TensorFlow Lite Micro, LSTM, and 1D-CNN models.
industrial-robotics-sensing-lab
CASPER real-time anomaly detection in industrial robotic arms using TinyML on Arduino Nicla Sense ME. Detects joint velocity deviations and external disruptions (earthquakes, impacts, load changes, magnetic interference) via LSTM/1D-CNN models deployed with TensorFlow Lite Micro, with OTA updates over BLE and RTDE integration for Universal Robots.