Inference Verification System Decimal Parameter Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing zero-knowledge proof methods can only handle integer values as parameters of inference models, limiting their applicability to decimal parameter models.
Innovation Solution
An inference verification system that expresses decimal values as integer values for use in convolutional neural networks, generates proofs using these values, and verifies the results using specific algorithms to handle decimal parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If zero-knowledge proof protocols are used for inference verification, then inference result authenticity is proven, but only integer parameter models can be handled
Solution Approach 1:
The patent transforms the parameter representation from decimal to integer by introducing a scaling factor. The inference model parameters are multiplied by a predetermined power of 10 to convert them into integers, which then can be processed by existing zero-knowledge proof protocols that only support integer arithmetic. This parameter transformation resolves the contradiction by maintaining reliability through cryptographic proof while expanding adaptability to support decimal parameter models.
2Reliability
If inference model parameters are converted to integers, then zero-knowledge proof verification becomes possible, but model precision may be affected
Solution Approach 1:
The patent applies parameter transformation by scaling decimal parameters into integers using a predetermined power of 10 as the scaling factor. This transformation enables verification capability while managing precision through the choice of scaling factor, which determines the resolution of the converted integer parameters.
Solution Approach 2:
The scaling factor acts as an intermediary between the original decimal parameter space and the integer parameter space required by zero-knowledge proof protocols. This intermediary transformation layer preserves the essential characteristics of the original parameters while making them compatible with verification protocols, thus maintaining measurement precision through reversible transformation.
3Loss of information
If the inference model is disclosed to the client, then verification transparency is achieved, but intellectual property is exposed
Solution Approach 1:
The patent extracts only the necessary verification information from the inference model through zero-knowledge proofs. Instead of disclosing the entire model, the system generates cryptographic proofs that verify the model's execution without revealing the model parameters themselves. This extraction approach achieves verification transparency while protecting intellectual property by separating verification needs from model disclosure.
Solution Approach 2:
Zero-knowledge proofs serve as an intermediary mechanism between the service provider and client. The proof system acts as a mediator that enables verification of inference results without requiring direct access to or disclosure of the underlying inference model, thus resolving the contradiction between transparency and IP protection.
Data Source
AI summary
An inference device (400) obtains an inference result by executing an inference model by expressing a decimal value that is data on which inference processing is to be performed as an integer value and treating the integer value as a parameter of a convolutional neural network. A proving device (500) obtains a proof by executing a proof generation algorithm using the inference result as input. A verification device (600) obtains a verification result by executing a verification algorithm using the proof as input.