AdvBox
AdvBox is a comprehensive toolbox designed to generate adversarial examples that can deceive neural networks across multiple frameworks, including PaddlePaddle, PyTorch, Caffe2, MxNet, Keras, and TensorFlow. It provides a command line tool that allows users to create adversarial examples with zero coding, making it accessible for researchers and developers alike.
Key Features:
- Multi-Framework Support: Works with popular machine learning frameworks such as PaddlePaddle, PyTorch, Caffe2, MxNet, Keras, and TensorFlow.
- Zero-Coding Tool: Users can generate adversarial examples without needing extensive programming knowledge.
- Robustness Benchmarking: Evaluate the robustness of machine learning models against adversarial attacks.
- Part of AdvBox Family: Includes various tools for adversarial example generation, detection, and protection, enhancing AI model security.
Benefits:
- Ease of Use: The command line interface simplifies the process of generating adversarial examples.
- Research and Development: Ideal for academic research and practical applications in AI security.
- Community Support: Backed by a community of contributors and users, ensuring continuous improvement and updates.
Highlights:
- Supports Python 3.*
- Inspired by FoolBox v1, ensuring a solid foundation for adversarial example generation.
- Open-source with an Apache License 2.0, promoting collaboration and transparency.