Machine Learning Trust Scoring for Untrusted User Equipment Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless communication systems face challenges in identifying and managing untrusted user equipment (UEs) that provide corrupted or perturbed data for machine learning operations, which can degrade the performance and security of the network.
Innovation Solution
A method and apparatus for wireless communications that utilize a machine learning model to determine the trustworthiness of data from UEs by performing outlier detection and assigning trust scores, allowing trusted data to be used for model training while excluding untrusted data, and potentially terminating connections or restricting services for untrusted UEs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data from all UEs is collected for machine learning training, then the quantity of training data increases, but the reliability of the machine learning model degrades due to corrupted or perturbed data from untrusted UEs
Solution Approach 1:
The patent extracts and removes untrusted or corrupted data from the training dataset by using machine learning models to identify and separate reliable data from unreliable data. This allows the system to maintain a large quantity of training data while eliminating the harmful portions that would degrade model reliability.
Solution Approach 2:
The patent introduces an intermediary machine learning model that acts as a filter between data collection and model training. This intermediary model evaluates the trustworthiness of incoming data and determines which data points should be included in training, thereby protecting the final model from corrupted data while preserving valuable training samples.
2Reliability
If outlier detection and trust scoring mechanisms are implemented, then the reliability of data selection improves, but the device complexity increases
Solution Approach 1:
The patent implements self-service by using machine learning models that automatically evaluate and score the trustworthiness of incoming data without requiring manual intervention. The system autonomously identifies outliers, assigns trust scores, and makes decisions about data inclusion, reducing the need for complex manual verification processes while maintaining high reliability.
3Object-affected harmful factors
If untrusted UEs are identified and restricted, then the security of the network improves, but the loss of information increases as potentially valid data from misclassified UEs is excluded
Solution Approach 1:
The patent uses parameter changes by dynamically adjusting trust scores and data inclusion thresholds based on the evaluation of each data point. Rather than using fixed binary classifications, the system continuously adjusts parameters like trust scores to reflect the actual quality of incoming data, allowing flexible inclusion or exclusion decisions that minimize information loss while maintaining security.
Data Source
AI summary
Methods, systems, and devices for wireless communications are described. In some systems, a network entity may obtain information (e.g., a data set, a model update) corresponding to a user equipment (UE), the information associated with a machine learning model. The network entity may determine whether the information or the UE providing the information is trusted or untrusted based on the information. The network entity may output, to another network entity, an indication that the information corresponding to the UE is considered untrusted or trusted based on a predicted output of the machine learning model (e.g., if the model is trained using the information). The other network entity may further train the machine learning model using trusted information and may refrain from using untrusted information. Additionally, or alternatively, if a UE is determined to be untrusted, a network entity may configure the UE to refrain from further data collection processes.