: Seduction and Vulnerability: Analyzing Human-Centric Exploitation in Media-Based Social Engineering.
Recent advancements in focus on "Alarm Token-Guided Safety" to help AI systems explicitly perceive and reason through harmful or deceptive video content [1]. This research aims to shift security from simply blocking files to understanding the context of why a file might be dangerous. SeducedAndVulnerable.mp4
: Moving from passive harm recognition to active defensive reasoning using frameworks like VideoSafety-R1 [1]. SeducedAndVulnerable.mp4
: Seduction and Vulnerability: Analyzing Human-Centric Exploitation in Media-Based Social Engineering.
Recent advancements in focus on "Alarm Token-Guided Safety" to help AI systems explicitly perceive and reason through harmful or deceptive video content [1]. This research aims to shift security from simply blocking files to understanding the context of why a file might be dangerous.
: Moving from passive harm recognition to active defensive reasoning using frameworks like VideoSafety-R1 [1].