Lightweight Detection of Denial-of-Service Attacks on Wireless Sensor Networks Revisited

The resource-constrained nature of sensor nodes makes wireless sensor networks (WSNs) especially susceptible to denial-of-service (DoS) attacks. Due to the wireless communication medium, it is difficult to prevent attacks such as jamming. Hence, mechanisms to detect attacks during operation are required. The current generation of intrusion detection systems are still rather heavyweight, as some form of collaboration is typically needed. In this paper, we study the behavior of a large number of node-centric metrics under jamming and blackhole attacks, by applying a logistic regression. In our experiments, we vary several parameters, such as traffic intensity, transmission power, and attacker location. We consider the most common topologies in wireless sensor networks such as central data collection and meshed multi-hop networks by using the collection tree and the mesh protocol. The created regression models are then used to implement a fully localized intrusion detection system requiring no collaboration, showing that certain models can be generalized to different networks.