[论文解读] 5G enabled Mobile Edge Computing security for Autonomous Vehicles

Daryll Ralph D'Costa, Hassan Abbas|arXiv (Cornell University)|Jan 30, 2022

Vehicular Ad Hoc Networks (VANETs)被引用 4

一句话总结

本文提出了一种基于机器学习的框架，用于检测和分类支持自动驾驶汽车的5G移动边缘计算（MEC）环境中的分布式拒绝服务（DDoS）攻击。该研究通过添加合成的5G网络参数来增强通用数据集，应用标签编码和少数类过采样技术，并训练模型以预测DDoS攻击和5G网络时延质量，但性能受限于真实5G数据不足以及对模拟单机实验的依赖。

ABSTRACT

The world is moving into a new era with the deployment of 5G communication infrastructure. Many new developments are deployed centred around this technology. One such advancement is 5G Vehicle to Everything communication. This technology can be used for applications such as driverless delivery of goods, immediate response to emergencies and improving traffic efficiency. The concept of Intelligent Transport Systems (ITS) is built around this system which is completely autonomous. This paper studies the Distributed Denial of Service (DDoS) attack carried out over a 5G network and analyses security attacks, particularly the DDoS attack. The aim is to implement a machine learning model capable of classifying different types of DDoS attacks and predicting the quality of 5G latency. The initial steps of implementation involved the synthetic addition of 5G parameters into the dataset. Subsequently, the data was label encoded, and minority classes were oversampled to match the other classes. Finally, the data was split as training and testing, and machine learning models were applied. Although the paper resulted in a model that predicted DDoS attacks, the dataset acquired significantly lacked 5G related information. Furthermore, the 5G classification model needed more modification. The research was based on largely quantitative research methods in a simulated environment. Hence, the biggest limitation of this research has been the lack of resources for data collection and sole reliance on online data sets. Ideally, a Vehicle to Everything (V2X) project would greatly benefit from an autonomous 5G enabled vehicle connected to a mobile edge cloud. However, this project was conducted solely online on a single PC which further limits the outcomes. Although the model underperformed, this paper can be used as a framework for future research in Intelligent Transport System development.

研究动机与目标

• 为解决自动驾驶汽车所依赖的5G移动边缘计算（MEC）环境中的安全漏洞，特别是DDoS攻击问题。
• 开发一种机器学习模型，以在5G网络环境中对不同类型DDoS攻击进行分类。
• 在MEC和自动驾驶汽车运行背景下，将5G网络时延质量作为性能指标进行预测。
• 评估使用合成5G参数扩充现有数据集以训练鲁棒DDoS检测模型的可行性。

提出的方法

• 将合成的5G网络参数添加到现有数据集中，以模拟真实世界的5G MEC环境。
• 数据预处理包括对分类特征进行标签编码，并对少数攻击类别进行过采样以平衡数据集。
• 采用标准划分技术将数据集划分为训练集和测试集。
• 训练并评估了多种机器学习模型，用于DDoS攻击分类和5G时延预测。
• 使用标准分类指标评估模型性能，尽管摘要中未报告具体数值。
• 研究完全在单台PC上的仿真环境中进行，依赖公开可用的数据集。

实验结果

研究问题

• RQ1 在支持自动驾驶汽车的5G移动边缘计算环境中，机器学习模型在分类不同类型DDoS攻击方面的有效性如何？
• RQ2 合成5G参数在多大程度上能提升现有数据集在车载网络DDoS检测中的代表性与实用性？
• RQ3 在基于MEC的智能交通系统中，机器学习模型能否在DDoS攻击条件下准确预测5G网络时延质量？
• RQ4 在真实自动驾驶汽车场景中，使用仿真单机环境和合成数据训练与验证DDoS检测模型存在哪些局限性？

主要发现

• 所提出的机器学习模型在模拟的5G MEC环境中成功检测并分类了DDoS攻击。
• 尽管模型功能正常，但性能受限于真实5G网络数据不足以及对合成参数的过度依赖。
• 本研究使用的数据集严重缺乏真实的5G相关信息，影响了模型的泛化能力和可靠性。
• 模型预测5G时延质量的能力也受到相同的数据限制和仿真约束的制约。
• 研究揭示了在5G赋能的自动驾驶汽车网络中，训练鲁棒DDoS检测系统面临真实世界数据严重不足的关键缺口。
• 尽管模型表现不佳，但该框架为智能交通系统和5G MEC安全领域的未来研究提供了基础性方法。