Unsupervised Learning Algorithms for Denial of Service Detection in Vehicular Networks

Authors:

Abstract
Denial of Service ( $D$ oS) attacks are a major threat for vehicular networks. Detecting and identifying the $D$ oS traffic is crucial for defending against such attacks. Machine Learning (ML) algorithms have been extensively adopted in traffic classification and detection of network attacks, namely the DoS attacks. Among and unlike different ML learning models, Unsupervised Learning (UL) algorithms have not being used in the literature for DoS detection. This paper shed the light on the feasibility of using unsupervised learning algorithms for detecting DoS attacks. It analyzes and compares the detection efficiency of selected UL algorithms using the Vehicular Reference Misbehavior (VeReMi) dataset [1]. Finally, simulation demonstrates the performance and efficiency of the used UL algorithms in $D$ oS detection; in particular, the Gaussian Mixture Model (GMM) algorithm demonstrates a detection accuracy with more than 95% for all $D$ oS attack traffic categories.

This publication has 19 references indexed in Scilit: