AI Model Assessment Provenance Using Hashes and Digital Signatures

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AI model assessments lack provable provenance, making it difficult to verify the authenticity of input data, models, and results due to the absence of reliable audit logs and susceptibility to tampering, which can obscure bias detection.

Innovation Solution

Implementing signed code and hash values for AI model assessments to ensure the integrity of the validation dataset and model, using a trusted party to sign the assessment framework, and generating hash values for the dataset and model to detect tampering, followed by verification using a public key infrastructure.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If audit logs are maintained to provide provenance information, then the reliability of assessment results is improved, but the device complexity and cost of infrastructure increase significantly

Engineering Contradiction:
Improveprovenance verificationVSAvoidinfrastructure complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts the essential provenance verification function from complex audit log infrastructure. Instead of maintaining comprehensive logs of all AI model operations, the system uses cryptographic hashing and digital signatures to verify only the critical provenance information (dataset, model, code, results), eliminating the need for complex logging infrastructure while maintaining reliability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent creates cryptographic copies (hash values and digital signatures) of the provenance information. These copies serve as tamper-evident records that can be verified without storing the entire audit log history, reducing infrastructure complexity while maintaining verification capability.

Inventive Principle:
Principle #26Copying

2Reliability

If audit logs are used to track modifications, then the reliability of assessment results is improved, but the ease of operation deteriorates due to difficulty in detecting and preventing tampering

Engineering Contradiction:
Improvetamper detectionVSAvoidtamper detection difficulty
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The patent implements a feedback mechanism where cryptographic hashes and digital signatures provide immediate verification of data integrity. When data is modified, the hash mismatch immediately signals tampering, eliminating the need to search through complex audit logs and making tamper detection straightforward and automatic.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces the mechanical/manual process of auditing and searching through log files with automated cryptographic verification. The digital signature and hash mechanisms automatically detect tampering without requiring manual investigation, significantly improving ease of operation while maintaining high reliability.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If comprehensive audit logging is implemented, then the reliability of provenance information is improved, but the loss of time increases due to the cost and maintenance requirements

Engineering Contradiction:
Improveprovenance trackingVSAvoidmaintenance time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts only the essential provenance verification functionality from comprehensive audit logging. By using cryptographic hashes and digital signatures on critical data elements, the system maintains provenance reliability without the time-consuming maintenance of extensive log infrastructure and processes.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4280543B1Systems and methods for provable provenance for artificial intelligence model assessments
Publication Date: 2025.09.03 CREDO AI CORP
  • EP4280543B1 patent drawingFigure 1
  • EP4280543B1 patent drawingFigure 2
  • EP4280543B1 patent drawingFigure 3

AI summary

Systems and methods are described herein for providing provable provenance for assessment results. For example, an AI model and/or a dataset may be assessed using an assessment service to determine whether a bias exists within the AI model and/or the dataset. The results of the assessment may be provided to an auditing service to confirm the assessment results. The systems and methods described herein provide for provable provenance for the assessment results such that the auditing service can verify whether a model and validation dataset provided by a client are the same that were used during an assessment and have not been tampered with by a malicious party.