Generating Realistic Datasets for VANET Misbehavior Detection
MSc Thesis Proposal by: Joshua Picchioni
Date: February 27th 2025
Time: 1:00 PM – 2:30 PM
Location: OBB04 (Odette)
Abstract:
Current VANET Misbehavior Detection benchmarks are typically created using simulated data and often present easily separable attacks, failing to challenge classifiers or produce generalized models. To address this, we are developing a framework that transforms real-world data from the Wyoming Connected Vehicle Pilot into high-difficulty experimental datasets. By embedding sophisticated attack patterns like Random and Constant Position Offsets directly into genuine traffic, we create subtle classification boundaries that are significantly harder to train on. We will evaluate this method using a suite of machine learning algorithms to demonstrate the increased difficulty, creating a new benchmark essential for challenging future machine learning models.
Keywords: VANETs, Misbehavior Detection, Machine Learning, Connected Vehicles, Wyoming CV Pilot
Thesis Committee:
Internal Reader: Dr. Jianguo Lu
External Reader: Dr. Bala Balasingam
Advisor: Dr. Arunita Jaekel
Chair: