Companies Borrow Attack Technique to Watermark Machine Learning Models
Researchers continue to improve on a technique for embedded crafted outputs into machine-learning models, an anti-copying technique originally thought up by adversarial researchers.
Source: darkreading.com