New attack model targets safety-critical video perception systems

Researchers have developed a novel framework for near-real-time targeted object removal attacks on safety-critical video-based perception systems. This attack model can compromise intelligent transportation systems by manipulating video frames, leading to failures in safety-critical functions. Experiments demonstrated that the reconstructed frames maintain high similarity to originals, with a significant reduction in object detections and a high attack success rate, indicating a vulnerability that could impact vision-based pedestrian safety systems. AI

IMPACT Highlights potential vulnerabilities in AI-powered safety systems, necessitating the development of robust defense mechanisms.

RANK_REASON Academic paper detailing a new attack model on video perception systems. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• Mohammad Imtiaz Hasan
• SC-CVT
• ScienceCast
• South Carolina Connected Vehicle Testbed
• YOLO

• safety
• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →