ai-evals

adversarial-robustness-toolbox

Name: adversarial-robustness-toolbox
Author: Trusted-AI

Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Teams

6,285 stars+8★ this week1,351 forksPythonUpdated 10/23/2026100% free · open source

What it does

A Python library that lets you test and improve your ML model’s security by generating and defending against adversarial attacks.

When to use it

•You have a production model and need to measure how vulnerable it is to evasion attacks before launch.
•Your security team wants to simulate poisoning or extraction attacks on a prototype to assess risk.
•You’re building a “red‑team” toolset for a data‑driven product and need ready‑made attacks and defenses.

Ready-to-paste prompt

python run_fgsm.py  # runs the script above, loading `mnist_cnn.h5`, attacking the test set, and printing the degraded accuracy

Heads up: ART requires the same deep‑learning framework version that your model was trained with (e.g., TensorFlow 2.x). Mismatched versions cause import errors, so verify `pip show tensorflow` matches the version used to build the saved model.

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

✅ Light setup

Installs with a command or two; your AI agent can do it for you.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/adversarial-robustness-toolbox && curl -fsSL https://workflowstacks.com/api/skills/adversarial-robustness-toolbox/claude-skill -o ~/.claude/skills/adversarial-robustness-toolbox/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How adversarial-robustness-toolbox works

Codeflow

Free to inspect

Adversarial-robustness-toolbox is a very large Python project (~103k lines across 396 code files, plus 239 test files). Setup is light: installs like a normal app. Reading the code is optional. Last commit 9 months ago, MIT license, has a test suite.

Size

Very large codebase

~103k lines · 396 code files · days to read — use, don't read

Setup

One-command install

Installs like a normal app. Reading the code is optional.

Runs on

Python · Docker

No API keys detected

Python 100%

Where to start reading

1
README.md
Start here — what it does and how to install it
2
pyproject.toml
Dependencies and the commands it exposes
3
examples/README.md
A worked example — copy this to get going

What's in each folder

docs/Documentation65 files

examples/Examples you can copy22 files

art/Folder361 files

utils/Helper scripts139 files

notebooks/Research & notebooks94 files

tests/Tests — proof it works251 files

.github/CI / automation (GitHub Actions)34 files

READMEHas testsDocumentedCI checksDocker readyExamples includedMIT licenseLast update 9 mo ago

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

✅ Light setup

Installs with a command or two; your AI agent can do it for you.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/adversarial-robustness-toolbox && curl -fsSL https://workflowstacks.com/api/skills/adversarial-robustness-toolbox/claude-skill -o ~/.claude/skills/adversarial-robustness-toolbox/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

Quick Actions

Details

Creator

Trusted-AI

Language

Python