An introduction to zero-knowledge machine learning (ZKML) (worldcoin.org)

• help us determine that a particular piece of content was produced by applying a specific model to a given input
• verifying outputs from large language models
• creating zero-knowledge proofs of the inference step of the ML model, not about the ML model training (which, in and of itself, is already very computationally intensive).

Application

• Computational integrity (validity ML)
  • prove and verify that the output is the product of a given model and input pair
• **ML as a Service (MLaaS) transparency**
  • it is really hard to know as a user whether the service provider is actually providing the model that they say they are since the API is a black box
• ZK anomaly/fraud detection
• Privacy (ZKML)

Papers

HE vs ZKML

• HE provides full data privacy by allowing developers to perform operations on encrypted data and decrypting the result to obtain the output of the operation on the original input. HE is used for full data privacy.
• ZKML requires the prover to have access to all the data, but the verifiers do not access the data. ZKML is used to create zero-knowledge proofs of the inference step of the ML model, while

Groth16

ZK Terminology