Echtzeit-Ml-Modellaktualisierung

Details

Veröffentlichungs-Nr.

EP4811216

Anmeldetag

21. März 2025

Veröffentlichung

23. September 2026

Rechtsraum

EP

Abstract

A real-time machine learning (ML) model updating method is described, which comprises monitoring input data to a deployed ML model, detecting drift of the input data over time, identifying new patterns in the input data, and responsive to detected drift, retraining the ML model using the identified new patterns. In some implementations, the ML model is deployed in a cyber-threat detection application operating on a computer network, wherein the input data comprises network metadata, and wherein the new patterns are indicative of new network threats. In this way, a need for real-time monitoring of ML models for detecting the changes in the behaviour of known threats and exposing the ML models to the newly identified traits of the threat can be provided to achieve a reliable threat detection system.

Anmelder

Firma

British Telecommunications Public Limited Company

Land

🇬🇧 Großbritannien