As more users ask AI systems for information, AI-search platforms are becoming a common gateway to web information. Unlike traditional search, which maps keywords to ranked pages, AI search retrieves pages, filters sources, selects citations, and generates answers before users see sources. This selection layer may ampl...
Qi Liu, Geng Hong, Xin-Yan Zhang et al.· 0 citations
A non-invasive model fingerprinting framework based on collapsed generation, a phenomenon where certain input conditions produce highly consistent images across multiple stochastic seeds, is presented, establishing collapsed generation as a reliable intrinsic evidence source for non-invasive diffusion model ownership v...
Yuanmin Huang, Chen Chen, Geng Hong et al.· 0 citations
Mobile agents powered by foundation models now automate complex, multi-step workflows on real devices, but their trajectories can violate app-specific security policies. Existing trajectory-level defenses rely on LLM prompting or rigid rules, and thus fail to support fine-grained, natural-language policies that general...
Chang-Yue Jiang, Jia-Yi Wang, Xin Wen et al.· 1 citation
This work provides a comprehensive study of web-based MSSO deployment and security implications, and designs a hierarchical detection framework using passive DNS correlations and URL reconstruction from search data to identify MSSO-enabled websites.
Jia-Sheng Huang, Ming-Xuan Liu, Pei Chen et al.· 0 citations
This study revisits how MCP security is measured and presents MCPZoo, the largest collection of MCP servers for dynamic analysis to date, and conducts the first ecosystem-scale measurement of MCP servers and the scanners that analyze them.
Pei Chen, Baichao An, Mengying Wu et al.· arXiv.org· 6 citations
This work proposes ActivationBackdoor, an inference-time backdoor attack that composes two activation-level components for trigger detection and backdoor behavior injection, and shows that ActivationBackdoor attains attack success comparable to training-time backdoor baselines while preserving high clean-task accuracy...
Zichun Su, Mi Zhang, Xiaohan Zhang et al.· Proceedings of the 32nd ACM...· 0 citations
This work proposes AEGIS (Adaptive Evasion Guard via Identification and Steering), an inference-time defense that applies similarity-aware repulsion only at the identified vulnerable heads of VSA and preserves benign fidelity, avoids suppressing hard-negative concepts, and transfers to SD 2.1 and FLUX.
Yuanmin Huang, Zhenfei Zhang, Mi Zhang et al.· arXiv.org· 0 citations
Collaborative inference enables cost-effective deployment of large language models by partitioning layers across multiple participants and forwarding intermediate activations between participants in a pipeline, but these transmitted activations also create a new attack surface: a malicious participant can manipulate in...
Zichun Su, Mi Zhang, Xiaohan Zhang et al.· Proceedings of the 32nd ACM...· 0 citations
Results show that evidence-grounded detection can assess both execution-related and semantic risks in MCP interactions, and compared with existing dynamic scanners, FlowGuard reduces end-to-end latency by up to 2.23x.
Baichao An, Pei Chen, Geng Hong et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
