MagDet:基于地磁的无人机GPS欺骗检测方法

MagDet:UAV GPS Spoofing Detection Based on the Geomagnetic Field

GPS是目前最为广泛使用的基于卫星的导航和定位系统,对于无人机而言,它是一个不可或缺的组成部分,提供了关键的精确位置数据,对导航和任务的成功至关重要.然而,GPS欺骗攻击已经逐渐演变成对GPS依赖系统不断增加的威胁.目前,针对无人机的GPS欺骗检测方法大多基于仿真数据提出、依赖于多个无人机或者需要专用设备(例如,软件定义无线电平台和高清摄像头).本文提出一种新颖的基于地磁场的无人机GPS欺骗检测方法--Mag Det(Magnetic field-based Detection method),其基本思想是利用地球内部和周围金属建筑材料不均匀性引起的地磁场异常,通过真实飞行收集位置和磁场强度数据,包括正常和被攻击场景.应用各种机器学习算法来训练这些数据以选择最佳分类器,该分类器可以轻松部署在常见机载计算机中.该方法的检测率超过99.5%,平均错误率(Equal Error Rate,EER)为0.51%,优于现有检测方法.此外,评估了各种因素对Mag Det的影响,以证明其鲁棒性.即使在未访问过的地点(距离6 km),准确率也高于95%,EER为0.49%.

GPS is currently the most widely used satellite-based navigation and positioning sys-tem.For unmanned aerial vehicles(UAVs),it is an indispensable component,providing crucial and precise location data that is essential for the success of navigation and missions.However,GPS spoofing attacks have gradually evolved into a growing threat to GPS-dependent systems.Most existing GPS spoofing detection methods for UAVs are proposed based on simulation data,and they depend on multiple UAVs or require dedicated devices(e.g.,software-defined radio platform and high-definition camera).In this paper,we propose a novel GPS spoofing detection framework,MagDet,for a UAV based on the geomagnetic field.Our basic idea is to use the geo-magnetic field anomalies due to inhomogeneities within the Earth's interior and surrounding met-al material.We collect positions and the strength of magnetic field through real flights including normal and attacked scenarios.Various machine learning algorithms are applied to train with these data to choose the best classifier,which can be easily deployed in common companion com-puter.The detection rate is more than 99.5%and the equal error rate(EER)is 0.51%,which is better than existing methods.We also evaluate various factors on MagDet to demonstrate the ro-bustness.Even in an unvisited site(6 kilometers away),the accuracy is higher than 95%and the EER is 0.49%.