An intrusion detection system for constrained WSN and IoT nodes based on binary logistic regression.

Abstract

In this article, we evaluate the feasibility of operating a lightweight intrusion detection system within a constrained sensor or IoT node. We propose mIDS, which uses statistical analysis tools based on binary logistic regression (BLR) to monitor and detect attacks. mIDS takes as input only local node parameters for both benign and malicious behavior and derives a normal behavior model that detects anomalies within the constrained node. We provide proof of correct operation by testing mIDS in the presence of network-level attacks. In such a system, critical data is obtained from the routing layer  and used as a basis for profiling the sensor behavior. Our results show that the proposed solution achieves attack detection accuracy ranging from 96% to 100% despite its lightweight implementation.