Intelligent Network Security Monitoring Based on Optimum-Path Forest Clustering

Abstract

Distinguishing outliers from normal data in wireless sensor networks has been a big challenge in the anomaly detection domain, mostly due to the nature of the anomalies, such as software or hardware failures, reading errors or malicious attacks, just to name a few. In this article, we introduce an anomaly detection-based OPF classifier in the aforementioned context. The results are compared against one-class support vector machines and multivariate Gaussian distribution. Additionally, we also propose to employ meta-heuristic optimization techniques to fine-tune the OPF classifier in the context of anomaly detection in wireless sensor networks.