Communications of the ACM
Researchers Develop Protocols to Validate Integrity of ML Models
Researchers at Texas A&M University (TAMU) and the University of California, Berkeley have validated the integrity of machine learning (ML) models, using cryptographic algorithms called zero-knowledge proof protocols. The protocols are a mathematical method that enables the owner of an ML model to generate a succinct proof to demonstrate with overwhelming probability that something is true without sharing extra data.
TAMU's Yupeng Zhang and colleagues designed new zero-knowledge proof methods and optimizations engineered to transform the computations of a decision-tree model into zero-knowledge proof statements. Zhang said, "These protocols will allow the owner of a machine learning model to prove to others that the model can achieve a high accuracy on public datasets without leaking any information about the machine learning model itself."
From Texas A&M University
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found