EAGER: SARE: Collaborative Research: Exploring and Mitigating Attacks of Millimeter-wave Radar Sensors in Autonomous Vehicles

EAGER：SARE：协作研究：探索和减轻自动驾驶汽车中毫米波雷达传感器的攻击

• 批准号: 2028872
• 负责人: Wenyao Xu
• 金额: $ 17万
• 依托单位: SUNY at Buffalo
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2028872&HistoricalAwards=false
• 关键词: EAGER; SARE; Collaborative; Research; Exploring

Autonomous vehicles (AVs) have significant potentials to improve road safety, travel experience and transportation footprint, as well as the mobility and accessibility for all. However, the safety and security of AVs themselves have raised many concerns in recent years. As AVs use a plethora of sensors, including cameras and radars to detect, classify and track objects and obstacles on the road, their reliability and resilience have significant impacts on AV safety. Although risks from malicious attacks targeted at cameras and other sensors in AV systems have been studied, there is a lack of in-depth understanding of the vulnerability of millimeter-wave radar sensors in AVs. This EArly-concept Grant for Exploratory Research (EAGER) project seeks to explore insights to improve the security and resilience of AVs by investigating advanced attack and defense methods for millimeter-wave radar sensor-based systems. Since these radars are used exclusively today for adaptive cruise control, blind-spot detection, and collision avoidance, this project will benefit many safety-critical applications. The PIs plan to expand research opportunities for K-12 and underrepresented students, and integrate diversity in broadening participation in engineering through the educational programs of National Summer Transportation Institute (NSTI), Institute for Sustainable Transportation and Logistics (ISTL) and Louis Stokes Alliance for Minority Participation (LSAMP) at the University at Buffalo. The PIs will disseminate the results of the project through publications, talks, and demos, and integrate research materials into specific courses and education curricula. All newly developed research and teaching materials will be publicly accessible on the project website. This project will first demonstrate that the millimeter-wave radar sensors can be spoofed and jammed while AVs are on the road through a non-cooperative over-the-air synchronization method. It can accurately identify the frequency band, modulation scheme and waveform patterns of victim radars to launch stealthy attacks. The team will conduct a proof-of-concept demo to attack real-world AV radars with fast-chirp signals. Second, the project will explore both hardware and software/algorithm-based defense mechanisms for avoiding such attacks, including beam feature based physical layer capacity estimation, machine learning physical identification, non-cooperative passive front-end architectures and band-limited coherent noise radars. The innovation of the project comes from not only the use of adaptive finite state machine based approaches that combine inter- and intra-chirp-sequence synchronization, but also data modeling techniques to efficiently adjust the attacker’s waveform parameters. Moreover, the project is among the first to explore the use of unique radiometrics, noise radars and passive radars as effective defense mechanisms against the attacks to AVs. Research thrusts in this project will significantly advance the state-of-the-art knowledge of the security of millimeter-wave sensors, and provide insights on developing more undeceivable, disclosure-resistant and robust AV radar solutions.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

自动驾驶汽车 (AV) 在改善道路安全、出行体验和交通足迹以及所有人的机动性和可达性方面具有巨大潜力，然而，近年来，自动驾驶汽车本身的安全性引起了许多担忧。大量的传感器，包括用于检测、分类和跟踪道路上的物体​​和障碍物的摄像头和雷达，其可靠性和弹性对自动驾驶汽车的安全性具有重大影响，尽管针对自动驾驶汽车系统中的摄像头和其他传感器的恶意攻击的风险已经被研究过。有一个对自动驾驶汽车中毫米波雷达传感器的脆弱性缺乏深入了解。这个早期概念探索性研究资助 (EAGER) 项目旨在通过研究先进的攻击和防御方法来探索提高自动驾驶汽车安全性和弹性的见解。由于这些雷达目前仅用于自适应巡航控制、盲点检测和防撞，因此该项目将有利于许多安全关键应用。 K-12 和代表性不足的学生，并通过大学国家夏季运输学院 (NSTI)、可持续运输和物流研究所 (ISTL) 和路易斯斯托克斯少数族裔参与联盟 (LSAMP) 的教育计划，将多样性融入扩大工程参与范围PI 将通过出版物、演讲和演示来传播该项目的成果，并将研究材料纳入具体​​的课程和教育课程中。该项目将首先演示毫米波雷达传感器可以在自动驾驶汽车行驶时通过非合作空中同步方法进行欺骗和干扰，它可以准确识别频段、调制方案和技术。该团队将进行概念验证演示，以利用快速线性调频信号攻击现实世界的 AV 雷达。用于避免此类攻击的基于软件/算法的防御机制，包括基于波束特征的物理层容量估计、机器学习物理识别、非合作无源前端架构和带限相干噪声雷达。该项目的创新来自于非。不仅使用基于自适应有限状态机的方法，结合了内部和内部线性调频序列同步，而且还使用数据建模技术来有效地调整攻击者的波形参数。此外，该项目是第一个探索使用独特的方法的项目之一。辐射测量、噪声雷达和无源雷达作为针对自动驾驶汽车攻击的有效防御机制，该项目的研究重点将显着提高毫米波传感器安全性的最新知识，并为开发更可靠的传感器提供见解。 、防泄露和强大的 AV 雷达解决方案。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力优点和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Towards Backdoor Attacks against LiDAR Object Detection in Autonomous Driving

• DOI: 10.1145/3560905.3568539
• 发表时间: 2022-11
• 期刊: Proceedings of the 20th ACM Conference on Embedded Networked Sensor Systems
• 作者: Yan Zhang;Yi Zhu;Zihao Liu;Cheng-yi Miao;Foad Hajiaghajani;Lu Su;Chunming Qiao