Collaborative Threshold Watermarking

Tameem Bakr, Anish Ambreth, Nils Lukas

In federated learning (FL), $K$ clients jointly train a model without sharing raw data. Because each participant invests data and compute, clients need mechanisms to later prove th…