RAILS

RAILS - Robust Adversarial Immune-inspired Learning System

Platform

• Python: 3.7
• PyTorch: 1.5.0

Intro

RAILS is a computational method to boost robustness of prototype robust DNNs. RAILS uses an Adaptive Immune System Emulator (AISE) for robustness.

HOWTO

Run rails_cifar.ipynb, rails_mnist.ipynb, associate files include aise.py, pgd.py

Commercial Application

RAILS has been filed as a patent. No commercial usage is allowed.

Refer to this Repo.

If you use this code, please cite the following reference

@article{wang2020rails,
  title={RAILS: A Robust Adversarial Immune-inspired Learning System},
  author={Wang, Ren and Chen, Tianqi and Lindsly, Stephen and Rehemtulla, Alnawaz and Hero, Alfred and Rajapakse, Indika},
  journal={arXiv preprint arXiv:2012.10485},
  year={2020}
}