7 papers
RATION: Towards Quantitative Risk-Based Access Control for AI Agents
Yuntao Du, Harsh Chaudhari, Yuetian Chen +8
AI agents increasingly interact with external environments to complete complex user tasks. These interactions can lead to unexpected behavior and expose agents to security threats…
BrowseSafe++: Benchmarking Prompt Injection Detection Across Web Contexts, Languages, and Browser Views
Yuetian Chen, Kaiyuan Zhang, Yuntao Du +6
Browser agents need to make decisions and execute actions on behalf of users while consuming untrusted webpages, creating opportunities for malicious instructions embedded in webpa…
Membership Inference Attacks Against Fine-tuned Diffusion Language Models
Yuetian Chen, Kaiyuan Zhang, Yuntao Du +5
Diffusion Language Models (DLMs) represent a promising alternative to autoregressive language models, using bidirectional masked token prediction. Yet their susceptibility to priva…
Window-based Membership Inference Attacks Against Fine-tuned Large Language Models
Yuetian Chen, Yuntao Du, Kaiyuan Zhang +4
Most membership inference attacks (MIAs) against Large Language Models (LLMs) rely on global signals, like average loss, to identify training data. This approach, however, dilutes…
Imitative Membership Inference Attack
Yuntao Du, Yuetian Chen, Hanshen Xiao +2
A Membership Inference Attack (MIA) assesses how much a target machine learning model reveals about its training data by determining whether specific query instances were part of t…
Cascading and Proxy Membership Inference Attacks
Yuntao Du, Jiacheng Li, Yuetian Chen +5
A Membership Inference Attack (MIA) assesses how much a trained machine learning model reveals about its training data by determining whether specific query instances were included…
SOFT: Selective Data Obfuscation for Protecting LLM Fine-tuning against Membership Inference Attacks
Kaiyuan Zhang, Siyuan Cheng, Hanxi Guo +8
Large language models (LLMs) have achieved remarkable success and are widely adopted for diverse applications. However, fine-tuning these models often involves private or sensitive…