Protected ML Model Provisioning in Wireless Network Analytics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for protecting machine learning models in 3GPP wireless communication networks leave them exposed to unauthorized access and attacks, risking data breaches and misuse.
Innovation Solution
Implementing a Network Data Analytics Function (NWDAF) with a Model Training logical function (MTLF) that generates a protected trained ML model using a stored security context and sends a response message with the protected model file and security context information, along with a data collector that stores and sends the protected model to an Analytics Data Repository Function (ADRF) for secure storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If ML models are stored in ADRF without protection, then ease of access and operation is improved, but security and reliability deteriorate due to unauthorized access and data breaches
Solution Approach 1:
The patent introduces an encryption intermediary layer between the ML models and the ADRF storage system. The encryption function acts as a mediator that transforms plaintext ML models into ciphertext before storage, and the decryption function acts as a mediator that transforms ciphertext back into plaintext for authorized access. This intermediary encryption/decryption mechanism enables both secure storage and controlled access, resolving the contradiction between ease of access and security.
2Reliability
If ML models are protected using encryption, then security and reliability are improved, but device complexity and processing overhead increase
Solution Approach 1:
The patent applies preliminary action by performing encryption on ML models before they are stored in the ADRF. The encryption process is executed in advance during the model provisioning phase, so that when models are retrieved for inference, they are already in protected form. This preliminary encryption action shifts the computational complexity to the model upload phase rather than the inference phase, reducing the burden on the inference system while maintaining security.
3Speed
If security context is stored locally in data collector, then access speed is improved, but security risk increases due to potential compromise of the data collector
Solution Approach 1:
The patent segments the security context storage and management across multiple entities rather than concentrating it in a single data collector. The encryption function and security context are separated from the data collector, and the ADRF stores encrypted models while the NWDAF manages security contexts. This segmentation distributes the security burden and reduces the attack surface, as compromising a single entity no longer provides access to all security contexts and models.
Data Source
AI summary
There is provided a method in a Network Data Analytics Function containing a Model Training logical function. The method comprises receiving a machine learning (ML) model provision request, the ML model provision request comprising: an identifier for at least one Analytic, and, ML model file specific information, and generating a protected trained ML model using a stored security context. The method further comprises sending, in response to the ML model provision request, an ML model provision response message, the ML model provision response message comprising: the identifier for the at least one Analytic; at least one protected trained ML model file; and location information of the stored security context.