Intelligent intrusion detection in external communication systems for autonomous vehicles
Intelligent intrusion detection in external communication systems for autonomous vehicles
复制标题
DOI：
10.1080/21642583.2018.1440260
复制
发表时间：
2018-01-01
影响因子：
4.1
通讯作者：
McDonald-Maier, Klaus
中科院分区：
文献类型：
--
作者：
Alheeti, Khattab M. Ali;McDonald-Maier, Klaus
众所周知，自动驾驶车辆由于这些车辆中使用的通信系统的类型而容易受到不同类型的攻击。这些车辆正变得越来越依赖于通过车辆自组织网络的外部通信。然而，这些网络为自动驾驶汽车带来了新的威胁，这可能会导致自动驾驶系统出现重大问题。这些通信系统可能会使自动驾驶车辆受到恶意攻击，如常见的Sybil攻击，黑洞，拒绝服务，虫洞攻击和灰洞攻击。本文提出了一种智能保护机制，该机制旨在保护自动驾驶和半自动汽车的外部通信。保护机制基于比例重叠分数方法，该方法允许减少在京都基准数据集中发现的特征的数量。这种混合检测系统使用反向传播神经网络检测拒绝服务（DoS），一种常见的攻击类型，在车载自组织网络。实验结果表明，该入侵检测方法能够识别自动驾驶甚至半自动驾驶车辆中的恶意车辆。
Self-driving vehicles are known to be vulnerable to different types of attacks due to the type of communication systems which are utilized in these vehicles. These vehicles are becoming more reliant on external communication through vehicular ad hoc networks. However, these networks contribute new threats to self-driving vehicles which lead to potentially significant problems in autonomous systems. These communication systems potentially open self-driving vehicles to malicious attacks like the common Sybil attacks, black hole, Denial of Service, wormhole attacks and grey hole attacks. In this paper, an intelligent protection mechanism is proposed, which was created to secure external communications for self-driving and semi-autonomous cars. The protection mechanism is based on the Proportional Overlapping Scores method, which allows to decrease the number of features found in the Kyoto benchmark dataset. This hybrid detection system uses Back Propagation neural networks to detect Denial of Service (DoS), a common type of attack in vehicular ad hoc networks. The results from our experiment revealed that the proposed intrusion detection has the ability to identify malicious vehicles in self-driving and even in semi-autonomous vehicles.