A new research paper details a method for near-real-time targeted object removal attacks on video-based perception systems used in intelligent transportation. The attack framework, tested on the South Carolina Connected Vehicle Testbed, successfully reduced object detections by up to 97.59% and achieved a 94.48% frame-level attack success rate. The reconstructed frames maintained high similarity to originals, making them difficult to distinguish, and the attack operated quickly enough for real-time application, posing a threat to safety-critical functions. AI
IMPACT Demonstrates a significant vulnerability in AI-powered safety systems, potentially impacting autonomous vehicle perception and road safety.
RANK_REASON The cluster contains a research paper detailing a novel attack model and framework. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
- Fast Object Removal Attacks on Safety-Critical Video-based Perception Systems
- Hugging Face Daily Papers
- South Carolina Connected Vehicle Testbed
- YOLO
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
