Vehicular ad-hoc networks (VANETs) present security vulnerabilities, which make them prone to diverse cyberattacks. Denial of Service (DoS) is one of the most prevalent and severe cyberattack that targets VANETs. To tackle this cyberattack and mitigate its effect, intrusion detection systems need to be developed. To this end, a realistic and representative dataset is essential to train and validate the systems. This paper proposes a new dataset, VDoS-LRS, which includes legitimate and simulated vehicular network traffic, along with different types of DoS cyberattack. We also present a realistic testbed environment instead of simulators, taking into consideration different environments (urban, highway and rural). In addition, we explore a wide range of traffic features for detecting and classifying vehicular traffic. We evaluate the reliability of the VDoS-LRS dataset using different machine learning algorithms for forensics purposes. The experimental results showed that it is possible to detect effectively different types of DoS cyberattack within diverse environments.

车载自组织网络（VANET）存在安全漏洞，使它们易于遭受各种网络攻击。拒绝服务（DoS）是针对VANET的最普遍，最严重的网络攻击之一。为了应对这种网络攻击并减轻其影响，需要开发入侵检测系统。为此，一个真实且具有代表性的数据集对于训练和验证系统至关重要。本文提出了一个新的数据集VDoS-LRS，其中包括合法和模拟的车辆网络流量以及不同类型的DoS网络攻击。考虑到不同的环境（城市，高速公路和乡村），我们还提供了一个逼真的测试平台环境，而不是模拟器。此外，我们探索了多种交通功能，以检测和分类车辆交通。为了进行取证，我们使用不同的机器学习算法评估了VDoS-LRS数据集的可靠性。实验结果表明，可以在各种环境中有效地检测不同类型的DoS网络攻击。