Abstract
Methods and systems for training or fine-tuning machine learning models (e.g., diffusion models) and generating zero-knowledge proofs are disclosed. Such zero-knowledge proofs can demonstrate that a machine learning model was trained or fine-tuned using specified data elements (e.g., provided by a client). A computer system (e.g., a server computer) can receive a request from a client computer to train a machine learning model using specified data elements. The server computer can use a hash function to generate noise corresponding to the specified data elements, which can be bound to the specified data elements. The computer system can train or fine-tune the machine learning model using a modified loss function. The computer system can generate an arithmetic circuit and a zero-knowledge proof based on the arithmetic circuit. The zero-knowledge proof can prove that the computer system did train or fine-tune the machine learning model using the specified data elements.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Bing-Jyue Chen; Bidgoli, Mohammad Mohsen Minaei; and Le, Duc, "PROOF OF FINETUNING DATA", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11098