LaserTag: A Tool for Autonomous XAI-Guided Physical Adversarial Perturbations in Industrial Vision Pipelines
Year: 2026
Type: software
Abstract: This repository implements a general closed-loop physical attack framework for vision-based monitoring systems. It performs autonomous, XAI-guided laser perturbations and includes a full suite of digital color-filter defenses. Manual attacks using a gamepad are also supported. The system is device-agnostic: adapting to a new device only requires replacing classifier weight files. For general information, see the main README.md file. See XAI-Auto-Laser-Attack-and-Defense-main/Siemens-Energy-T3000/README.md for a detailed description of a experiment including specific requirements (hardware + sofware) to reproduce results.
Institution: Karlsruhe Institute of Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7140681229