Safe and Robust Watermark Injection with a Single OoD Image
A new method for safely and robustly injecting watermark after training without training data.
Who Leaked the Model? Tracking IP Infringers in Accountable Federated Learning
Tracking IP leakage in federated learning.