Gelei Deng

Key contribution not extracted.

Mobile GUI agents are surprisingly susceptible to prompt injection via realistic, attacker-controlled text embedded within ordinary user-generated content, even without modifying the agent, application, or OS.

VideoLLMs leak training data: a novel black-box attack recovers membership with surprisingly high accuracy (AUC=0.68) by probing generation brittleness across temperatures.

Automatically exploiting web application vulnerabilities is now significantly more feasible, with AutoEG achieving over 80% success where previous methods struggled to reach 33%.

Humans are surprisingly vulnerable to deception by compromised LLM agents, with less than 10% detecting attacks even in high-stakes scenarios.

LLMs and LVLMs share more than half their top-activated neurons during multi-step inference, opening a surprisingly cheap path to boost vision-language reasoning by transplanting skills from text-only models.

LLM-powered pentesting agents fail not because of model limitations, but because they can't estimate task difficulty, leading to wasted effort and premature context exhaustion.