# From Component Manipulation to System Compromise: Understanding and Detecting Malicious MCP Servers

arXiv: https://arxiv.org/abs/2604.01905 · HTML: https://mlsanity.com/paper/2604.01905/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 2 Apr 2026.

## Built on by (most cited first, 4 of 4)

- [When Safe Skills Collide: Measuring Compositional Risk in Agent Skill Ecosystems](https://mlsanity.com/paper/2606.00448/) (2026)
- [Behavioral Integrity Verification for AI Agent Skills](https://mlsanity.com/paper/2605.11770/) (2026)
- [Red-Teaming Agent Execution Contexts: Open-World Security Evaluation on OpenClaw](https://mlsanity.com/paper/2605.11047/) (2026)
- [WebMCP Tool Surface Poisoning: Runtime Manipulation Attacks on LLM Agents](https://mlsanity.com/paper/2606.06387/) (2026)

## Abstract (arXiv)

The model context protocol (MCP) standardizes how LLMs connect to external tools and data sources, enabling faster integration but introducing new attack vectors. Despite the growing adoption of MCP, existing MCP security studies classify attacks by their observable effects, obscuring how attacks behave across different MCP server components and overlooking multi-component attack chains. Meanwhile, existing defenses are less effective when facing multi-component attacks or previously unknown malicious behaviors. This work presents a component-centric perspective for understanding and detecting malicious MCP servers. First, we build the first component-centric PoC dataset of 114 malicious MCP servers where attacks are achieved as manipulation over MCP components and their compositions. We evaluate these attacks' effectiveness across two MCP hosts and five LLMs, and uncover that (1) component position shapes attack success rate; and (2) multi-component compositions often outperform single-component attacks by distributing malicious logic. Second, we propose and implement Connor, a two-stage behavioral deviation detector for malicious MCP servers. It first performs pre-execution analysis to detect malicious shell commands and extract each tool's function intent, and then conducts step-wise in-execution analysis to trace each tool's behavioral trajectories and detect deviations from its function intent. Evaluation on our curated dataset indicates that Connor achieves an F1-score of 94.6%, outperforming the state of the art by 8.9% to 59.6%. In real-world detection, Connor identifies two malicious servers.