Date of this Version
N. Jaton, “Distributed neural network based architecture for DDoS detection in vehicular communication systems,” M.S. Thesis, Univ. Nebraska-Lincoln, 2021.
With the continued development of modern vehicular communication systems, there is an ever growing need for cutting edge security in these systems. A misbehavior detection systems (MDS) is a tool developed to determine if a vehicle is being attacked so that the vehicle can take steps to mitigate harm from the attacker. Some attacks such as distributed denial of service (DDoS) attacks are a concern for vehicular communication systems. During a DDoS attack, multiple nodes are used to flood the target with an overwhelming amount of communication packets. In this thesis, we investigated the current MDS literature and how it is used to prevent DDoS attacks. Additionally, we proposed and developed a new distributed multilayer perceptron classifier (MLPC) and evaluated it using vehicular communication simulations generated using OMNeT++, Veins, and Sumo. These simulations contained a group of normal vehicles and some attacking vehicles. During the simulations, two different attacks were conducted. Apache Spark was then used to create the distributed MLPC. The median F1-score for this MLPC architecture was 95%. This architecture outperformed linear regression and support vector machines, which achieved 89% and 88% respectively, but was unable to better random forests and gradient boosted trees which both achieved a 97% F1-score. Using Amazon Web Services (AWS), it was determined that model training and detection time were not significantly increased with the inclusion of additional nodes after three nodes including the master.
Advisor: Yi Qian