Hiroto Onoda, Go Tsuruoka, Yuna Tanaka, and Ryunosuke Kobayashi, Waseda University; Kento Oonishi, Mitsubishi Electric; Takuya Higashi, Mitsubishi Electric Corporation; Yoshihiro Koseki, Mitsubishi Electric; Tsunato Nakai, Mitsubishi Electric Corporation; Tatsuya Mori, Waseda University/NICT/RIKEN AIP
The accuracy of object detection technology is critical to automated driving safety. The poster introduces a new remote patch attack that applies a full-scene adversarial texture and defeats ObjectSeeker, the current state-of-the-art defence. Simulation tests show person-detection recall dropping by 43 points on YOLOv5s (VOC) and by 35 points even with ObjectSeeker active, while physical experiments on a 70″ monitor confirm the effect, preventing detection in 70 % of video frames.
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
