基于激光雷达的点云目标检测安全性综述  

Review on the safety of LIDAR-based point cloud object detection

作　　者：项圣 廖丹芯 陈晋音[1,2] 郑海斌 XIANG Sheng;LIAO Danxin;CHEN Jinyin;ZHENG Haibin(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China;Institute of Cyberspace Security,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023 [2]浙江工业大学网络安全研究院,浙江杭州310023

出　　处：《网络与信息安全学报》2025年第1期20-39,共20页Chinese Journal of Network and Information Security

基　　金：浙江省自然科学基金(LDQ23F020001);国家自然科学基金(62072406,62406286);国家重点研发计划(2018AAA0100801)。

摘　　要：激光雷达技术在自动驾驶、智能交通和环境感知等领域有着广泛应用。在自动驾驶系统中,利用机载摄像头和激光雷达等对周围环境进行感知,点云目标检测技术在环境感知方面发挥着重要的作用。近年来的研究指出,基于激光雷达的点云目标检测系统容易遭受欺骗攻击。然而,目前对基于激光雷达的点云目标检测的安全性问题的研究相对零散,设计的安全威胁场景、攻击方法和防御策略相互独立。基于现有研究工作,首次对基于激光雷达的点云目标检测的安全性进行了综述,主要关注该技术在实际应用中可能面临的安全挑战和解决方案。首先,全面回顾了基于激光雷达的目标检测和攻防理论;其次,从数字域和物理域双重视角分析了攻击方法,并对现有的防御策略进行了详尽的分类和总结;接着,介绍了该领域目前常用的数据集及评价标准;最后,结合面临的挑战和机遇,对基于激光雷达的目标检测安全的未来研究方向进行了总结和展望,为基于激光雷达的点云目标检测的安全研究的发展和应用提供参考。LiDAR technology has been widely applied in fields such as autonomous driving,intelligent traffic,and environmental perception.In autonomous driving systems,the surrounding environment was perceived using onboard cameras and LiDAR,with point cloud object detection playing a crucial role.Recent studies have indicated that point cloud object detection based on LiDAR was susceptible to deception attacks.However,research on the security issues of LiDAR-based object detection has been relatively scattered,with security threat scenarios,attack methods,and defense strategies being addressed independently.Based on existing research,the first comprehensive review of the security of LiDAR-based point cloud object detection was presented,focusing on the security challenges and solutions that the technology may face in practical applications.The theory of object detection and the principles of attack and defense based on LiDAR were comprehensively reviewed.Attack methods were analyzed from both the digital and physical domains,and existing defense strategies were classified and summarized in detail.Then,the commonly used datasets and evaluation criteria in this field were introduced.And considering the challenges and opportunities,a summary and outlook on future research directions in the security of LiDAR-based detection were provided,offering valuable insights for the development and application of secure research in LiDAR-based point cloud object detection.

关 键 词：激光雷达 目标检测 攻击方法 防御方法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]