|
Work item:
|
TR.lzkml
|
|
Subject/title:
|
Technical Report: Landscape analysis of Zero-Knowledge machine learning
|
|
Status:
|
Under study
|
|
Approval process:
|
Agreement
|
|
Type of work item:
|
Technical report
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2026-Q3 (Medium priority)
|
|
Liaison:
|
-
|
|
Supporting members:
|
-
|
|
Summary:
|
Machine learning (ML) has been widely applied across various fields, but ensuring the security and PII of ML models and data remains a significant challenge. Particularly, in decentralized and PII-sensitive environments, it is essential to verify the correctness of ML computations without exposing the underlying data or model details. Zero-Knowledge Machine Learning (ZKML) leverages Zero-Knowledge Proofs (ZKP) to enable PII-preserving and verifiable ML computations. By utilizing cryptographic proofs, ZKML allows a model owner to prove that an ML model’s inference or training process has been executed correctly without revealing sensitive information. This approach enhances trust in AI models, prevents adversarial manipulations, and ensures secure model verification. This Technical Report provides an overview of the current landscape of ZKML, highlighting its key use cases, security threats. It aims to serve as a reference for ITU-T members and researchers by practical applications of ZKML and assessing its potential for future standardization and widespread deployment.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-04-17 15:06:53
|
|
Last update:
2025-04-17 15:11:13
|
