Skip to content

A2M: Trace-Optimized Agent Hijacking in the MCP Ecosystem

Sep 2026 · Volume 1 · Vol abs/2609.26761 · 0 citations · 24 references
Computer Science

TL;DR

A2M (Attraction-to-Manipulation), a two-stage black-box framework for hijacking MCP agents, is introduced, a two-stage black-box framework for hijacking MCP agents that motivate stronger tool vetting and runtime isolation in MCP ecosystems.

Abstract

Agents using the Model Context Protocol (MCP) rely on semantic matching to select tools from third-party servers, exposing a semantic supply-chain risk through attacker-controlled metadata and outputs. We introduce A2M (Attraction-to-Manipulation), a two-stage black-box framework for hijacking MCP agents. The Attraction phase optimizes tool metadata to increase invocation probability; the Manipulation phase uses execution traces to refine adversarial tool returns that steer agents toward attacker-desired outcomes. On LiveMCPBench, direct attacks optimized and evaluated on GLM-4.6 achieve a macro-average malicious tool invocation rate of 93.6% across four scenarios, increase weighted token costs to 32.4$\times$ the benign baseline under Cognitive Denial of Service, and attain a mean attack success rate of 74.4% across Information Exfiltration, Environment Integrity Compromise, and Reasoning Derailment. Transfer to four other models without re-optimization yields corresponding macro-averages of 63.6%, 2.7$\times$, and 24.5%. These findings motivate stronger tool vetting and runtime isolation in MCP ecosystems. Code is publicly available at https://github.com/Lilaizhen/A2M.

View source

Similar papers

Preprint Aug 2026

TrustShiftProbe: Characterizing, Benchmarking, and Defending Staged Trust Attacks on MCP Servers

TrustShiftProbe is introduced, an evaluation and defense framework with four contributions: a stateful temporal threat model of the agent-server lifecycle as a benign conditioning phase followed by an adversarial defection at a trust horizon, and a language-agnostic attack engine that instantiates each variant as a com...

Mehrdad Rostamzadeh, Sidhant Narula, Mohammad GhasemiGol et al. · 1 citation
Preprint Aug 2026

Beyond Direct Access: Resource Hijacking in LLM Agents

This work is the first to identify and systematically study agent resource hijacking, a security blind spot in which attackers induce agents to invoke, consume, transfer, or control high-value resources for their own goals without directly obtaining those resources or their credentials.

Pu-Yu Zeng, Min-Gang Chen, Zhe-Li Liu et al. · 0 citations
Conference Open access Jun 2026

AEGIS: Preventing Cross-Domain Resource Abuse in MCP

AEGIS is presented, a policy enforcement component that enables administrators to define fine-grained safeguards against resource abuse across heterogeneous MCP tools and modalities and detects and mitigates abusive behaviors while preserving the flexibility of MCP-based agent ecosystems.

S. Priya, Teryl Taylor, Frederico Araujo · 1 citation
#artificial intelligence Preprint Oct 2026

COPEX: Benchmarking LLM Robustness to Adversarial Context Across Model Context Protocol Layers

Large language models increasingly mediate tool use in Model Context Protocol (MCP) systems, where adversarial influence may enter through user instructions, tool schemas, tool outputs, or protocol messages. Existing benchmarks often evaluate deployed agents, conflating model susceptibility with guardrails, orchestrati...

Nahom Birhan, Mehrdad Rostamzadeh, Sidhant Narula et al. · 0 citations
#artificial intelligence Preprint Oct 2026

EvoRiskBench: An Evolving Benchmark for Runtime Security Risks in Workspace Agents

Workspace agents combine large language models with execution harnesses to perform stateful, multi-step tasks that access or modify external resources. Existing benchmarks leave gaps in executable coverage of their runtime security risks, while evolving model capabilities, harnesses, tools, and threats motivate benchma...

Shi Kuang, Xue-Mei Luo, Kun Liu et al. · 0 citations
Preprint Oct 2026

Compromise Is Not Consequence: Evaluating Task-Scoped Authorization in LLM Agents with Paired Replay

A tool-using model can follow a malicious instruction even when its credentials are valid. We study whether task-scoped authorization contains the resulting tool execution. Our paired-replay testbed samples a model request once and submits the same action, resource, and arguments to broad bearer, scoped JWT, sender-con...

Tural Hagverdiyev · 1 citation

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.