Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AdvGPS: Adversarial GPS for Multi-Agent Perception Attack

Jinlong Li, Baolu Li, Xinyu Liu, Jianwu Fang, Felix Juefei-Xu, Qing Guo, Hongkai Yu

TL;DR

ADVGPS is introduced, a method capable of generating adversarial GPS signals which are also stealthy for individual agents within the system, significantly reducing object detection accuracy and substantially undermining the performance of state-of-the-art methods.

Abstract

The multi-agent perception system collects visual data from sensors located on various agents and leverages their relative poses determined by GPS signals to effectively fuse information, mitigating the limitations of single-agent sensing, such as occlusion. However, the precision of GPS signals can be influenced by a range of factors, including wireless transmission and obstructions like buildings. Given the pivotal role of GPS signals in perception fusion and the potential for various interference, it becomes imperative to investigate whether specific GPS signals can easily mislead the multi-agent perception system. To address this concern, we frame the task as an adversarial attack challenge and introduce \textsc{AdvGPS}, a method capable of generating adversarial GPS signals which are also stealthy for individual agents within the system, significantly reducing object detection accuracy. To enhance the success rates of these attacks in a black-box scenario, we introduce three types of statistically sensitive natural discrepancies: appearance-based discrepancy, distribution-based discrepancy, and task-aware discrepancy. Our extensive experiments on the OPV2V dataset demonstrate that these attacks substantially undermine the performance of state-of-the-art methods, showcasing remarkable transferability across different point cloud based 3D detection systems. This alarming revelation underscores the pressing need to address security implications within multi-agent perception systems, thereby underscoring a critical area of research.

AdvGPS: Adversarial GPS for Multi-Agent Perception Attack

TL;DR

ADVGPS is introduced, a method capable of generating adversarial GPS signals which are also stealthy for individual agents within the system, significantly reducing object detection accuracy and substantially undermining the performance of state-of-the-art methods.

Abstract

The multi-agent perception system collects visual data from sensors located on various agents and leverages their relative poses determined by GPS signals to effectively fuse information, mitigating the limitations of single-agent sensing, such as occlusion. However, the precision of GPS signals can be influenced by a range of factors, including wireless transmission and obstructions like buildings. Given the pivotal role of GPS signals in perception fusion and the potential for various interference, it becomes imperative to investigate whether specific GPS signals can easily mislead the multi-agent perception system. To address this concern, we frame the task as an adversarial attack challenge and introduce \textsc{AdvGPS}, a method capable of generating adversarial GPS signals which are also stealthy for individual agents within the system, significantly reducing object detection accuracy. To enhance the success rates of these attacks in a black-box scenario, we introduce three types of statistically sensitive natural discrepancies: appearance-based discrepancy, distribution-based discrepancy, and task-aware discrepancy. Our extensive experiments on the OPV2V dataset demonstrate that these attacks substantially undermine the performance of state-of-the-art methods, showcasing remarkable transferability across different point cloud based 3D detection systems. This alarming revelation underscores the pressing need to address security implications within multi-agent perception systems, thereby underscoring a critical area of research.
Paper Structure (13 sections, 8 equations, 4 figures, 3 tables)

This paper contains 13 sections, 8 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. V2V Perception Pipeline and Motivation
  4. AdvGPS against Multi-agent Perception
  5. Overview
  6. Appearance-based Discrepancy
  7. Distribution-based Discrepancy
  8. Task-aware Discrepancy
  9. Implementation Details
  10. Experiments
  11. Experimental Setups
  12. Quantitative Evaluation
  13. Conclusions

Figures (4)

  • Figure 1: Illustration of AdvGPS for multi-agent perception attack. Here we use Vehicle-to-Vehicle (V2V) cooperative perception in autonomous driving as an example. Ego vehicle might receive the shared visual information from other CAVs with the adversarial GPS signal, leading to significant false-negative and false-positive detection errors.
  • Figure 2: Illustration of V2V cooperative perception pipeline with LiDAR data coordinate projection from CAV to Ego.
  • Figure 3: Pipeline of AdvGPS for multi-agent cooperative perception attack. Dash lines indicate the back propagation.
  • Figure 4: 3D detection visualization on attacking V2V model CoBEVT xu2022cobevt. Blue and red 3D bounding boxes represent ground truth and prediction. Purple arrows: detection errors. Green point cloud: by ego vehicle. Other-color point clouds: projected to ego coordinate by nearby attacked CAVs.