Detection of Adversarial Attacks in Robotic Perception

TL;DR

This paper addresses the challenge of detecting adversarial attacks in robotic perception systems that use deep neural networks for semantic segmentation, emphasizing the need for specialized detection methods.

Contribution

It highlights the necessity for tailored detection strategies for adversarial attacks in robotic semantic segmentation, extending robustness research beyond image classification.

Findings

Semantic segmentation in robotics is vulnerable to adversarial attacks.

Existing robustness methods for classification are insufficient for segmentation.

Specialized detection strategies are needed for robotic perception systems.

Abstract

Deep Neural Networks (DNNs) achieve strong performance in semantic segmentation for robotic perception but remain vulnerable to adversarial attacks, threatening safety-critical applications. While robustness has been studied for image classification, semantic segmentation in robotic contexts requires specialized architectures and detection strategies.