AI Model Training via Licensed Operator Data Access

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In the context of Open Radio Access Networks (O-RAN) for 5G wireless networks, there is a challenge in providing vendors with sufficient and relevant operator data for training Machine Learning (ML) models, leading to sub-optimal solutions being deployed, as vendors often lack the necessary data volume, veracity, and variability for network operation use cases.

Innovation Solution

A method and apparatus that enables a training application to access operator data by requesting a license from a license management function, receiving a license credential, and using this credential to access operator data through a data management function, allowing the training of AI/ML models with the received data, while ensuring secure and controlled access.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If vendors train ML models with limited available data, then model training can proceed, but the model quality and performance deteriorate due to insufficient data volume, veracity, and variability

Engineering Contradiction:
Improvemodel qualityVSAvoiddata volume
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The patent introduces a data access intermediary system consisting of a data access request module, data access control module, and data platform. This intermediary architecture enables vendors to access operator data through controlled interfaces, resolving the contradiction by providing a mechanism to obtain sufficient training data while maintaining security and control. The intermediary system validates vendor credentials, checks data access permissions, and facilitates controlled data sharing between operators and vendors.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If operators share data freely with vendors, then vendors can train effective models, but operator data security and control deteriorate

Engineering Contradiction:
Improvemodel effectivenessVSAvoiddata security risk
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The data access control module acts as an intermediary that enforces security policies while enabling data sharing. It validates vendor credentials, checks access permissions against operator-defined policies, and controls data export through approved channels. This resolves the contradiction by maintaining operator control and security while still providing vendors with sufficient data for effective model training.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements differential access rights for different vendors, data types, and usage scenarios. The data access control module evaluates each data access request individually, applying specific security policies based on vendor credentials, data sensitivity, and intended use. This allows operators to share appropriate data with vendors while protecting sensitive information, resolving the contradiction between data sharing and security.

Inventive Principle:
Principle #3Local quality

3Object-affected harmful factors

If a license management system is implemented to control data access, then data security improves, but system complexity increases due to additional authentication and authorization components

Engineering Contradiction:
Improvedata securityVSAvoidsystem complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The data access control module performs multiple functions within a single component: credential validation, permission checking, policy enforcement, and data export control. The data platform also serves multiple purposes including data storage, access logging, and vendor management. This multi-functionality reduces overall system complexity while maintaining comprehensive security controls.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent combines the data access request module, data access control module, and data platform into an integrated system. The credential validation, permission checking, and data delivery functions are merged into a coordinated workflow that reduces the number of separate components needed. This consolidation maintains security while reducing system complexity compared to having separate systems for each function.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4099191A1Training of ai / ML based models with operator data
Publication Date: 2022.12.07 NOKIA TECHNOLOGIES OY
  • EP4099191A1 patent drawingFigure 1
  • EP4099191A1 patent drawingFigure 2
  • EP4099191A1 patent drawingFigure 3A

AI summary

A training application (TrApp) is configured to train an AI/ML based model; sending to a license management function (LiMnF) in a network a request (R1) for acquiring a license for accessing operator data generated by one or more network entities in the network; receive, from the license management function (LiMnF) in response to the request (R1) for acquiring a license, at least one license credential, wherein the license credential identifies operator data to which the training application (TrApp) is allowed to access; send at least one request (R3) for accessing operator data, the at least one request for accessing operator data including the at least one license credential; receive in response to the at least one request (R3) for accessing operator data the operator data to which the training application (TrApp) is allowed to access. The AI/ML based model may be trained using received operator data as training data.