Zhiqiang Wang

Large Language Models (LLMs) are evolving into autonomous agents capable of executing complex workflows via standardized protocols (e.g., MCP). However, this paradigm shifts control from deterministic code to probabilistic inference, creating a fundamental Trust-Authorization Mismatch: static permissions are structurally decoupled from the agent's fluctuating runtime trustworthiness. In this Systematization of Knowledge (SoK), we survey more than 200 representative papers to categorize the emerging landscape of agent security. We propose the Belief-Intention-Permission (B-I-P) framework as a unifying formal lens. By decomposing agent execution into three distinct stages-Belief Formation, Intent Generation, and Permission Grant-we demonstrate that diverse threats, from prompt injection to tool poisoning, share a common root cause: the desynchronization between dynamic trust states and static authorization boundaries. Using the B-I-P lens, we systematically map existing attacks and defenses and identify critical gaps where current mechanisms fail to bridge this mismatch. Finally, we outline a research agenda for shifting from static Role-Based Access Control (RBAC) to dynamic, risk-adaptive authorization.

To standardize interactions between LLM-based agents and their environments, the Model Context Protocol (MCP) was proposed and has since been widely adopted. However, integrating external tools expands the attack surface, exposing agents to tool poisoning attacks. In such attacks, malicious instructions embedded in tool metadata are injected into the agent context during MCP registration phase, thereby manipulating agent behavior. Prior work primarily focuses on explicit tool poisoning or relied on manually crafted poisoned tools. In contrast, we focus on a particularly stealthy variant: implicit tool poisoning, where the poisoned tool itself remains uninvoked. Instead, the instructions embedded in the tool metadata induce the agent to invoke a legitimate but high-privilege tool to perform malicious operations. We propose MCP-ITP, the first automated and adaptive framework for implicit tool poisoning within the MCP ecosystem. MCP-ITP formulates poisoned tool generation as a black-box optimization problem and employs an iterative optimization strategy that leverages feedback from both an evaluation LLM and a detection LLM to maximize Attack Success Rate (ASR) while evading current detection mechanisms. Experimental results on the MCPTox dataset across 12 LLM agents demonstrate that MCP-ITP consistently outperforms the manually crafted baseline, achieving up to 84.2% ASR while suppressing the Malicious Tool Detection Rate (MDR) to as low as 0.3%.

The evolution of Large Language Models (LLMs) into Agentic AI has established the Model Context Protocol (MCP) as the standard for connecting reasoning engines with external tools. Although this decoupled architecture fosters modularity, it simultaneously shatters the traditional trust boundary. We uncover a novel privacy vector inherent to this paradigm: the Intent Inversion Attack. We show that semi-honest third-party MCP servers can accurately reconstruct users' underlying intents by leveraging only authorized metadata (e.g., function signatures, arguments, and receipts), effectively bypassing the need for raw query access. To quantify this threat, we introduce IntentMiner. Unlike statistical approaches, IntentMiner employs a hierarchical semantic parsing strategy that performs step-level intent reconstruction by analyzing tool functions, parameter entities, and result feedback in an orthogonal manner. Experiments on the ToolACE benchmark reveal that IntentMiner achieves a semantic alignment of over 85% with original queries, substantially surpassing LLM baselines. This work exposes a critical endogenous vulnerability: without semantic obfuscation, executing functions requires the transparency of intent, thereby challenging the privacy foundations of next-generation AI agents.