Source: O'Reilly Media, Responsible Machine Learning, Patrick Hall, Navdeep Gill, and Benjamin Cox.
See secure_ml_ideas.pdf for a drafty write-up of these attacks.
Follow-ups to the original draft include:
- Proposals for model vulnerability and security on O'Reilly Ideas
- Can Your Machine Learning Model Be Hacked?! on the h2o.ai blog site.
See main.pdf for AI Risk and Security (AIRS) meetup slides.