Research map / Multi-agent systems and games / LLM agents
SOPE: Situation-Aware and Statistically Indistinguishable Privacy Exfiltration for MCP-enabled Agents
Ruixiao Lin, Qingming Li, Jiahao Chen, Chunyi Zhou, Shouling Ji
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes SOPE, a situation-aware, statistically indistinguishable privacy exfiltration attack against MCP-enabled LLM agents that exploits malicious tool supply chains.
Problem
Privacy leakage in MCP-enabled LLM agents via indirect prompt injection; existing attacks are misaligned with the agent's tool-usage context and are easily detectable.
Approach
Crafts exfiltration payloads that are aligned with the agent's situational tool-usage context and statistically indistinguishable from benign behavior, enabling covert privacy exfiltration through MCP tool supply chains.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the trained Assign module (ablation) | SOPE · Evasion against retrieval screening | ESR: 0.44 -> 0.87 | Improved |
| add the trained Assign module (ablation) | SOPE · Targeted, benchmark agents | leakage rate: 0.78 -> 0.92 | Improved |
| add the trained Assign module (ablation) | SOPE · Targeted, benchmark agents | ASR (total): 0.77 -> 0.75 | Hurt |
| replace frontier-LLM binary query with trained Assign (ablation) | SOPE · Targeted, benchmark agents | leakage rate: 0.62 -> 0.92 | Improved |
| add optimization-based Fuse instead of appending probing prompts (ablation) | SOPE · Targeted, benchmark agents | ASR (total): 0.23 -> 0.75 | Improved |
| add optimization-based Fuse instead of LLM rewriting (ablation) | SOPE · Targeted, benchmark agents | ASR (total): 0.41 -> 0.75 | Improved |
| add code-level Leak (transmit) instead of explicit instruction (ablation) | SOPE · Targeted, benchmark agents | ASR(S2): 0.46 -> 0.86 | Improved |
| add code-level Leak (transmit) instead of explicit instruction (ablation) | SOPE · Targeted, benchmark agents | ASR (total): 0.39 -> 0.75 | Improved |
| deploy IsolateGPT system-isolation defense against SOPE | SOPE · Targeted, benchmark agents | leakage rate: 0.92 -> 0.83 | Hurt |
| deploy retrieval-screening defenses against SOPE | SOPE · Targeted, benchmark agents | ASR (total): 0.75 -> 0.88 | Improved |
| target commercial agents instead of benchmark agents | SOPE · Targeted attack | leakage rate: 0.92 -> 0.94 | Improved |
| use SOPE instead of indirect-prompt-injection baselines | baseline attacks · Commercial agents, average | ASR: 0.35 -> 0.96 | Improved |
Experimental setup
Builds on: Model Context Protocol
Closest papers by meaning
- Searching for Privacy Risks in LLM Agents via Simulation (ICLR 2026 · similarity 0.86)
- Contextualized Privacy Defense for LLM Agents (ICML 2026 · similarity 0.86)
- PrivAct: Internalizing Contextual Privacy Preservation via Multi-Agent Preference Training (ICML 2026 · similarity 0.86)
- MCPHunt: An Evaluation Framework for Cross-Boundary Data Propagation in Multi-Server MCP Agents (NeurIPS 2026 · similarity 0.85)
- MINIM: Privacy-Aware Minimal View for Agents via Trusted Local Sanitization (ICML 2026 · similarity 0.85)
- PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in Action (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
