Tutorials & Examples Overview

The Tutorials & Examples section provides end‑to‑end, runnable scripts that show how to use AdversarialAttacks.jl in realistic workflows. Each tutorial is generated from the scripts in examples/ using Literate.jl, so the code runs during the documentation build and all output is captured inline.

The current tutorials cover:

• White-box FGSM attack against a small Flux CNN on MNIST — see the tutorial.
• Black-box Basic Random Search (SimBA-style) attack against a DecisionTree classifier on the Iris dataset — see the tutorial.
• Black-box SimBA attack on 2D spirals with decision boundary visualization — see the tutorial.

Running locally

The tutorial scripts in examples/ are fully standalone. You can run them directly from the repository root:

julia> using Pkg
julia> Pkg.activate("./examples")   # activate the examples environment from the repository root
julia> include("examples/whitebox_fgsm_flux_mnist.jl")
julia> include("examples/blackbox_basicrandomsearch_decisiontree_iris.jl")

Or run them directly from the examples/ directory using the project flag:

julia --project=. -e 'using Pkg; Pkg.instantiate(); Pkg.precompile()'
julia --project=. whitebox_fgsm_flux_mnist.jl
julia --project=. blackbox_basicrandomsearch_decisiontree_iris.jl

Notes and tips

• Each example uses a small training setup so it runs quickly for demonstration purposes. For a more applicable example, re-run with more epochs and save the trained model.
• Scripts will print training and attack statistics and generate plot files.
• If you want to reproduce results exactly, set seeds as shown in the examples.

Next steps

• Follow the individual tutorial pages for step‑by‑step code with executed output and plots