Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

MalTool: Malicious Tool Attacks on LLM Agents

About

In a malicious tool attack, an attacker uploads a malicious tool to a distribution platform; once a user installs the tool and the LLM agent selects it during task execution, the tool can compromise the user's security and privacy. Prior work primarily focuses on manipulating tool names and descriptions to increase the likelihood of installation by users and selection by LLM agents. However, a successful attack also requires embedding malicious behaviors in the tool's code implementation, which remains largely unexplored. In this work, we bridge this gap by presenting the first systematic study of malicious tool code implementations. We first propose a taxonomy of malicious tool behaviors based on the confidentiality-integrity-availability triad, tailored to LLM-agent settings. To investigate the severity of the risks posed by attackers exploiting coding LLMs to automatically generate malicious tools, we develop MalTool, a coding-LLM-based framework that synthesizes tools exhibiting specified malicious behaviors, either as standalone tools or embedded within otherwise benign implementations. To ensure functional correctness and structural diversity, MalTool leverages an automated verifier that validates whether generated tools exhibit the intended malicious behaviors and differ sufficiently from prior instances, iteratively refining generations until success. Our evaluation demonstrates that MalTool is highly effective even when coding LLMs are safety-aligned. Using MalTool, we construct two datasets of malicious tools: 1,200 standalone malicious tools and 5,287 real-world tools with embedded malicious behaviors. We further show that existing detection methods, including commercial malware detection approaches such as VirusTotal and methods tailored to the LLM-agent setting, exhibit limited effectiveness at detecting the malicious tools, highlighting an urgent need for new defenses.

Yuepeng Hu, Yuqi Jia, Mengyuan Li, Dawn Song, Neil Gong• 2026

Related benchmarks

TaskDatasetResultRank
Malicious Tool DetectionDataset II Trojan malicious tools 1.0 (test)
FNR0.061
60
Remote Data Exfiltration DetectionDataset I standalone malicious tools 1.0
FNR0.04
10
API Key Abuse DetectionDataset I standalone malicious tools 1.0
FNR0.00e+0
5
Environment Credential Harvesting DetectionDataset I standalone malicious tools 1.0
FNR0.43
5
Local Data Exfiltration DetectionDataset I standalone malicious tools 1.0
FNR71
5
Remote Program Downloading DetectionDataset I standalone malicious tools 1.0
FNR5
5
CPU Compute Hijacking DetectionDataset I standalone malicious tools 1.0
FNR96
5
Database Record Deletion DetectionDataset I standalone malicious tools 1.0
FNR69
5
GPU Compute Hijacking DetectionDataset I standalone malicious tools 1.0
FNR93
5
Local File Deletion DetectionDataset I standalone malicious tools 1.0
FNR50
5
Showing 10 of 12 rows

Other info

Follow for update