Distributed Device-Specific Anomaly Detection for Resource-Constrained Devices

The Internet of Things (IoT) requires security mechanisms that account for heterogeneous devices and resource constraints. Current Anomaly Detection (AD) approaches typically apply centralized data processing. This lacks scalability as well as privacy, bandwidth consumption, latency, and availability. The combination of device-specific AD models and distributed edge computing provides a solution to these challenges. It allows the creation of simplified, lightweight AD models that run on constrained hardware. The hands-on environment described allows interactive exploration of the impact of model simplification. This includes live evaluation of performance metrics and processing latency on a constrained device.