ML Model Key Lifecycle in NWDAF and ADRF Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current wireless communications systems lack a mechanism for managing security keys used to protect machine learning models, leading to potential security vulnerabilities and inefficient use of storage resources.
Innovation Solution
Implement a core network architecture with a Network Data Analytics Function (NWDAF) containing a Model Training Logical Function (MTLF) and an Analytics Data Repository Function (ADRF) to generate and manage security contexts, such as encryption and integrity protection keys, with defined storage and validity times to ensure timely deletion and refresh of these keys.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If security contexts are stored indefinitely to ensure continuous access to protected ML models, then system reliability is improved, but security vulnerabilities increase due to prolonged key exposure
Solution Approach 1:
The patent implements dynamic key management where security contexts are automatically refreshed after a validity period. The system transitions from static indefinite storage to dynamic time-limited storage with automatic renewal, balancing reliability and security by ensuring keys remain active for continuous operation while limiting exposure time through periodic regeneration
Solution Approach 2:
The patent introduces a validity time parameter for security contexts that changes the storage characteristics from permanent to temporary. By setting and enforcing a validity duration, the system modifies the storage parameter to automatically delete expired contexts, thereby reducing security vulnerabilities while maintaining operational reliability through controlled key lifecycle management
2Duration of action of stationary object
If protected ML models are stored indefinitely to ensure continuous availability, then service continuity is improved, but storage resource efficiency deteriorates
Solution Approach 1:
The patent implements periodic evaluation and automatic deletion of protected ML models based on their validity time. Instead of continuous indefinite storage, the system periodically checks validity periods and removes expired models, achieving resource efficiency through regular cleanup while maintaining service continuity for models within their valid period
Solution Approach 2:
The patent enables automatic discarding of expired protected ML models based on validity time expiration. The system recovers storage resources by deleting models that have completed their service lifecycle, thereby improving storage efficiency while maintaining service continuity for actively valid models through automated lifecycle management
3Reliability
If multiple security contexts are maintained for different ML models to ensure security, then security coverage is improved, but device complexity increases
Solution Approach 1:
The patent implements self-service automated key management where the system automatically generates, stores, validates, and deletes security contexts without manual intervention. The automated validity time checking and deletion mechanism reduces operational complexity while maintaining comprehensive security coverage across multiple ML models through systematic lifecycle management
4Productivity
If security contexts have extended validity periods to reduce refresh operations, then operational efficiency is improved, but security risk increases due to prolonged key exposure
Solution Approach 1:
The patent optimizes the validity time parameter to achieve the right balance between operational efficiency and security risk. By setting an appropriate validity duration that is neither too short nor too long, the system reduces the frequency of refresh operations improving efficiency, while simultaneously limiting key exposure time to maintain acceptable security risk levels
Data Source
AI summary
Various aspects of the present disclosure relate to a wireless communications system that includes a network data analytics function (NWDAF) containing a model training logical function (MTLF), an NWDAF containing an analytics logical function (AnLF), and an analytics data repository function (ADRF). The NWDAF containing the MTLF generates a security context that protects a machine learning (ML) model that is stored in the ADRF. An NWDAF containing the AnLF obtains the protected ML model from the ADRF and obtains the security context from the NWDAF containing the MTLF. The security context is managed using a storage duration time that indicates when the ADRF is to delete the protected ML and the NWDAF containing the MTLF is to delete the security context, or a validity time that indicates when the ADRF is to delete the protected ML and the NWDAF containing the MTLF is to delete the security context.