NWDAF Model Provisioning with Reliability Requirements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In current network data analytics systems, consumer Network Data Analytics Functions (NWDAFs) lack the ability to determine the accuracy of machine learning models provided by producer NWDAFs, which can lead to unreliable analytics due to insufficient user consent for model training.
Innovation Solution
The proposed solution involves allowing consumer NWDAFs to include a reliability requirement in model provisioning requests, specifying the required accuracy of the machine learning model. Producer NWDAFs then determine if they can provide a model meeting this requirement, and if not, they may reject the request or provide an alternative model with indicated accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If consumer NWDAF requests machine learning models from producer NWDAF without specifying reliability requirements, then model provisioning is simple and fast, but the accuracy and reliability of analytics cannot be guaranteed
Solution Approach 1:
The patent introduces a reliability requirement parameter in the model provisioning request message. This parameter allows the consumer NWDAF to specify the desired accuracy level (e.g., 90% confidence) for the requested analytics. The producer NWDAF uses this parameter to determine whether to fulfill the request, thereby enabling reliability control without fundamentally changing the provisioning mechanism.
Solution Approach 2:
The patent implements a feedback mechanism where the producer NWDAF responds to reliability requirements by either providing models that meet the specified accuracy threshold or rejecting the request with a cause indicator. This feedback loop enables the consumer to understand whether the requested analytics reliability can be achieved, allowing for informed decision-making without adding complex negotiation protocols.
2Productivity
If producer NWDAF collects user data for model training without checking user consent, then data collection is efficient and comprehensive, but user privacy is violated and model legitimacy is compromised
Solution Approach 1:
The patent requires the producer NWDAF to check user consent status in the UDM database before collecting user data for model training. This preliminary verification ensures that only data from users who have explicitly consented to analytics processing is collected, maintaining both efficiency (by filtering at the source) and legitimacy (by respecting user privacy preferences).
Solution Approach 2:
The patent uses the UDM (Unified Data Management) system as an intermediary to manage and verify user consent. The UDM stores subscription information including user consent preferences, and the producer NWDAF queries this intermediary to determine whether data collection is permitted. This intermediary approach centralizes consent management without adding complexity to the data collection process itself.
3Measurement precision
If consumer NWDAF accepts any machine learning model from producer NWDAF, then model provisioning is straightforward, but the accuracy of analytics cannot be verified or guaranteed
Solution Approach 1:
The patent adds a reliability requirement parameter to the model provisioning request that allows the consumer NWDAF to specify the minimum acceptable accuracy level for analytics. The producer NWDAF evaluates available models against this parameter and only provides models that meet the threshold, enabling accuracy control through a simple parameter specification rather than complex verification procedures.
Data Source
AI summary
When the consumer NWDAF requests a trained machine learning model from a producer NWDAF for a set of Analytic IDs associated with a plurality of UEs, the consumer NWDAF may include a reliability requirement in the model provisioning request to indicate a required accuracy for the machine learning model. The reliability requirement may be expressed in terms of a number of UEs, a percentage of UEs. or an accuracy target. The producer NWDAF determines whether it can provide a trained model satisfying the reliability requirement and responds accordingly. If a trained model meeting the reliability requirement is available, the producer NWDAF provides the location the trained model to the consumer NWDAF.