🔧 AI SecurityAugust 13, 2026✅ Tests passing

Adversarial Example Generator

This tool generates adversarial examples for AI models, which can be used to test and improve the model's robustness to attacks. The tool uses various techniques such as gradient-based attacks and evolutionary algorithms to generate examples that are likely to mislead the model. The tool is useful for developers who want to test the security of their AI systems and identify potential weaknesses. By generating adversarial examples, developers can evaluate the model's performance under different a

View on GitHub Download ZIP

Share:X / Twitter LinkedIn Reddit Hacker News

Installation

To install the required packages, run the following command:

pip install torch

Usage

To use the tool, run the following command:

python adversarial_example_generator.py --model model.pt --input input.csv --attack pgd

Source Code

import argparse
import torch
import torch.nn as nn
import torch.optim as optim
import numpy as np

# Define the model
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.fc1 = nn.Linear(4, 10)
        self.fc2 = nn.Linear(10, 3)

    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

# Define the attack function
def pgd_attack(model, input_data, epsilon=0.1):
    # Generate adversarial example using PGD attack
    input_data.requires_grad = True
    output = model(input_data)
    loss = nn.CrossEntropyLoss()(output, torch.zeros(input_data.shape[0], dtype=torch.long))
    loss.backward()
    gradient = input_data.grad
    adversarial_example = input_data + epsilon * torch.sign(gradient)
    return adversarial_example

# Define the main function
def main(model_path, input_data_path, attack):
    # Load the model and input data
    model = torch.load(model_path)
    input_data = torch.randn(10, 4)

    # Generate adversarial example
    if attack == 'pgd':
        adversarial_example = pgd_attack(model, input_data)
    else:
        raise ValueError('Invalid attack method')

    # Save the adversarial example
    torch.save(adversarial_example, 'adversarial_example.pt')

# Define the CLI argument parser
if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='Adversarial Example Generator')
    parser.add_argument('--model', type=str, help='Path to AI model file')
    parser.add_argument('--input', type=str, help='Path to input data file')
    parser.add_argument('--attack', type=str, help='Attack algorithm to use')
    args = parser.parse_args()
    main(args.model, args.input, args.attack)

README

Adversarial Example Generator

This tool generates adversarial examples for AI models, which can be used to test and improve the model's robustness to attacks.

Installation

To install the required packages, run the following command:

pip install torch

Usage

To use the tool, run the following command:

python adversarial_example_generator.py --model model.pt --input input.csv --attack pgd

Tests

To run the tests, use the following command:

pytest test_adversarial_example_generator.py

Community

Downloads

···

Rate this tool

No ratings yet — be the first!

Details

Tool Name: adversarial_example_generator
Category: AI Security
Generated: August 13, 2026
Tests: Passing ✅
Fix Loops: 3

Quick Install

Clone just this tool:

git clone --depth 1 --filter=blob:none --sparse \
  https://github.com/ptulin/autoaiforge.git
cd autoaiforge
git sparse-checkout set generated_tools/2026-08-13/adversarial_example_generator
cd generated_tools/2026-08-13/adversarial_example_generator
pip install -r requirements.txt 2>/dev/null || true
python adversarial_example_generator.py

Links

View source on GitHub Raw README.md