Let's make FAME famous! 🤩

While ML has been exceeding expectations across the board, it is time to take its safety concerns more serious.

The main goal is to evaluate and help drive improvements to ML models using open-source approaches.

FAME focus on identifying vulnerabilities in Windows malware classifiers by automatically searching for adversarial examples using multiple strategies:

• Stochastic Approach
• Genetic Programming
• Reinforcement Learning
• Universal Adversarial Perturbations

Moreover, it proposes a defense against adversarial attacks that aims to increase the cost of generating universal perturbations against a target malware classifier.

PRs are more than welcome! 💯