Agent Tools Orchestration Leaks More: Dataset, Benchmark, and Mitigation

arXiv:2512.16310 · cs.CR, cs.AI, cs.CL · Submitted 2025-12-18 · Read on arXiv

cs.CR, cs.AI, cs.CL

Submitted: 2025-12-18

Updated: 2026-09-02

Comments: Accepted to EMNLP 2026 Findings. 20 pages, 2 figures. Code and data: https://github.com/1Ponder/TOP-R

Code: https://github.com/1Ponder/TOP-R

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 32× in this paper · explained in 2 episodes
• Trace — 14× in this paper · explained in 2 episodes
• Core — 9× in this paper · explained in 2 episodes
• Grounding — 9× in this paper · explained in 3 episodes
• Compositional Tool Depth — 5× in this paper
• Provenance — 5× in this paper · explained in 3 episodes
• Leakage Rate — 4× in this paper
• Tools Orchestration Privacy Risk — 4× in this paper
• Benchmarking — 3× in this paper · explained in 3 episodes
• Calibration — 3× in this paper · explained in 7 episodes
• Reasoning Traces — 3× in this paper · explained in 3 episodes
• Thinking mode — 3× in this paper · explained in 2 episodes
• Behavioral and Activity Logs — 2× in this paper
• Completion Rate — 2× in this paper
• Context-Bounded Disclosure — 2× in this paper
• Financial Asset Data — 2× in this paper
• Health and Medical Data — 2× in this paper
• Library-Grounded Reverse-Inference Seed Expansion — 2× in this paper
• Proprietary and Confidential Information — 2× in this paper
• Reinforcement Learning — 2× in this paper · explained in 4 episodes
• Task-Necessary Minimality — 2× in this paper
• Abstraction — 1× in this paper · explained in 1 episode
• Agent Interactions — 1× in this paper · explained in 1 episode
• Aggregation — 1× in this paper · explained in 1 episode
• Alignment — 1× in this paper · explained in 1 episode
• Consistency — 1× in this paper · explained in 3 episodes
• Elias — 1× in this paper · explained in 2 episodes
• Failure Modes — 1× in this paper · explained in 3 episodes
• Multi-Agent System — 1× in this paper · explained in 5 episodes
• Multi-agent Collaboration — 1× in this paper · explained in 2 episodes

Sources

• CI-Work: Benchmarking Contextual Integrity in Enterprise LLM Agents
• MAGPIE: A benchmark for Multi-AGent contextual PrIvacy Evaluation
• CI-Bench: Benchmarking Contextual Integrity of AI Assistants on Synthetic Data
• Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory
• Safe RLHF: Safe Reinforcement Learning from Human Feedback
• Position: Privacy Is Not Just Memorization!
• CIMemories: A Compositional Benchmark for Contextual Integrity of Persistent Memory in LLMs
• The Sum Leaks More Than Its Parts: Compositional Privacy Risks and Mitigations in Multi-Agent Collaboration
• Privacy in Action: Towards Realistic Privacy Mitigation and Evaluation for LLM-Powered Agents
• The Rise and Potential of Large Language Model Based Agents: A Survey
• ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs
• SPILLage: Agentic Oversharing on the Web
• AgentLeak: A Benchmark for Internal-Channel Privacy Leakage in Multi-Agent LLM Systems
• Identifying the Risks of LM Agents with an LM-Emulated Sandbox
• InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents
• Evaluating Language Model Reasoning about Confidential Information
• Searching for Privacy Risks in LLM Agents via Simulation
• AgentDAM: Privacy Leakage Evaluation for Autonomous Web Agents
• Beyond Memorization: Violating Privacy Via Inference with Large Language Models

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs