Cisco’s AI security researchers have analyzed ways to target vision-language models (VLMs) using pixel-level perturbation.
The post Attackers Could Exploit AI Vision Models Using Imperceptible Image Changes appeared first on SecurityWeek.