USPTO Patent US12585730B2: Machine Learning Model Management

ChangeBridge: Patent Grants - AI & Computing (G06N)

Published March 24th, 2026

Detected March 25th, 2026

Summary

The USPTO has granted patent US12585730B2 to OPSWAT Inc. for a method of managing machine learning models used in file analysis. The patent describes a system that retrains models based on accuracy feedback to improve detection of suspicious files.

View original document View source feed page

What changed

The United States Patent and Trademark Office (USPTO) has granted patent US12585730B2, titled 'Managing machine learning models,' to OPSWAT Inc. The patent details a method for managing machine learning models, specifically for analyzing files to identify suspicious content. The system involves receiving files, deploying models to generate results, ranking and clustering these results, and calculating classes to distinguish accurate from inaccurate results. Crucially, the system retrains deployed models with new training sets when inaccuracies are detected and validates the retrained models by comparing their performance against the original models.

This patent grant is primarily an intellectual property matter and does not impose direct regulatory obligations on companies. However, it signifies a recognized innovation in the field of AI and machine learning for cybersecurity applications. Companies developing or utilizing similar machine learning-based threat detection systems should be aware of this patent, particularly concerning potential infringement if their systems employ comparable methodologies for model retraining and validation.

Source document (simplified)

← USPTO Patent Grants

Managing machine learning models

Grant US12585730B2 Kind: B2 Mar 24, 2026

Assignee

OPSWAT Inc.

Inventors

Ran Dubin

Abstract

A method includes receiving files. Deployed models analyze the files to create results with corresponding data. The results identify suspicious files. The system calculates ranked results by ranking the results according to a confidence level, and clustered groups by clustering the results based on similar data. The system calculates classes by correlating the ranked results to the clustered groups. The classes indicate accurate results and inaccurate results. The system retrains the deployed models having the class indicating inaccurate results with new training sets to create retrained models. The system validates the retrained models when the retrained models identify one or more suspicious files from the receiving. The system compares the classes of the deployed models to new classes of the retrained models. The computerized system deploys the retrained models when the new classes of the retrained models indicate accurate results greater than the classes of the deployed models.