#1414th of 14 in Security
adversarial-robustness-toolbox
Python library for attacking and defending ML models against adversarial threats
- Stars
- 6.3k
- License
- MIT
- Last commit
- Oct 2026
- Last release
- Jul 2025
- Language
- Python
5,873 on 4 Apr 20266,295 on 31 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
ART is a Python library for evaluating and hardening machine learning models against evasion, poisoning, extraction and inference attacks. It wraps models from TensorFlow, Keras, PyTorch, scikit-learn, XGBoost, LightGBM, CatBoost and GPy behind common estimator interfaces. It covers images, tables, audio and video, and is hosted by LF AI & Data.
Who it is for: ML engineers and researchers testing model robustness
Strengths
- Covers four threat classes: evasion, poisoning, extraction and inference
- Works across TensorFlow, PyTorch, scikit-learn, XGBoost, LightGBM, CatBoost and GPy
- Supports images, tables, audio and video, plus detection and speech tasks
- MIT license, LF AI & Data project, with examples and notebooks
Weaknesses
- README does not mention LLM-specific red-teaming; focus is classic model robustness
- Last release (2025-07-07) is older than recent commits
- Library only, no UI or hosted service; you write Python to use it
- Main docs sit on a GitHub wiki, with API reference on Read the Docs
What it needs
- GPU optional
- Docker
- Models: TensorFlow, Keras, PyTorch, scikit-learn, XGBoost
Quick answers
Is adversarial-robustness-toolbox open source?
Yes. Its code is on GitHub under the MIT license.
Can you run adversarial-robustness-toolbox with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Does adversarial-robustness-toolbox need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in Security
See all 14| Rank | Project | Score |
|---|---|---|
| 1 | shannonAutonomous pentester that exploits web apps and APIs using source code analysis | 91 out of 100 |
| 2 | ifixaiPython CLI that audits AI agents for policy, access and reliability failures | 86 out of 100 |
| 3 | StrixAutonomous AI pentesting agents that validate vulnerabilities with working exploits | 84 out of 100 |
| #4 | pentagiAutonomous multi-agent penetration testing system with sandboxed security tools | 78 out of 100 |
| #5 | ai-infra-guardRed-teaming platform for AI infra CVEs, MCP servers, agent skills and jailbreaks | 71 out of 100 |