ZK ML Proof
ZK Proof - proofs whether the content was produced by a neural network. Such approach is useful in various fields, for instance, generative art or plagiarism checking.
Created on 12th November 2022
•
ZK ML Proof
ZK Proof - proofs whether the content was produced by a neural network. Such approach is useful in various fields, for instance, generative art or plagiarism checking.
The problem ZK ML Proof solves
The solution is similar to the detection of AI-generated data, e.g., images or text. ZK approach allows us to publicly verify whether the provided data was generated by a specific neural network. Moreover, service providers (e.g., OpenAI) can provide ZK proofs along with generated outputs so everybody can make sure it's original.
The possible use cases;
- Detect if an image was generated using a generative model (Dalle-2, Midjourney, etc.).
- Check if the text was altered by "anti-plagitarism" neural network.
- Check if service provider (e.g., OpenAI) provides the output from the neural network you have selected.
Challenges we ran into
The problem area is complex, it requires deep understanding of cryptography and ZK proofs. Since our team aims on ML-related things, we lacked time to implement the proposed concept.
Discussion
Builders also viewed
See more projects on Devfolio