17818315. DATA SIGNATURES FOR ML SECURITY simplified abstract (QUALCOMM Incorporated)

From WikiPatents
Jump to navigation Jump to search

DATA SIGNATURES FOR ML SECURITY

Organization Name

QUALCOMM Incorporated

Inventor(s)

Mohamed Fouad Ahmed Marzban of San Diego CA (US)

Wooseok Nam of San Diego CA (US)

Tao Luo of San Diego CA (US)

Mahmoud Taherzadeh Boroujeni of San Diego CA (US)

DATA SIGNATURES FOR ML SECURITY - A simplified explanation of the abstract

This abstract first appeared for US patent application 17818315 titled 'DATA SIGNATURES FOR ML SECURITY

Simplified Explanation

The patent application describes a method for training a machine learning (ML) model using multiple datasets collected by different user equipment (UE) devices. Each dataset includes a set of metrics collected by a UE. The method involves assigning a data signature to each dataset, which represents the source of the dataset.

The innovation involves identifying a corrupted dataset by detecting a first data signature associated with it. The method then filters out the datasets associated with the first data signature from the training process of the ML model. This ensures that the corrupted dataset does not negatively impact the training of the ML model.

  • The patent application describes a method for training an ML model using multiple datasets collected by UEs.
  • Each dataset includes metrics collected by a UE, and a data signature is assigned to each dataset.
  • A corrupted dataset is identified by detecting a specific data signature associated with it.
  • Datasets associated with the corrupted data signature are filtered out from the training process of the ML model.

Potential Applications

  • This technology can be applied in various fields where ML models are trained using datasets collected from multiple sources.
  • It can be used in telecommunications networks to train ML models for network optimization or anomaly detection.
  • It can be applied in healthcare for training ML models using patient data collected from different medical devices.

Problems Solved

  • The method solves the problem of corrupted datasets negatively impacting the training of ML models.
  • It helps in ensuring the quality and reliability of the training data by filtering out corrupted datasets.
  • It improves the accuracy and performance of ML models by excluding datasets associated with corrupted data signatures.

Benefits

  • The method improves the overall quality of ML models by filtering out corrupted datasets.
  • It saves computational resources by excluding datasets that may introduce errors or biases in the training process.
  • It enhances the reliability and trustworthiness of ML models by ensuring the integrity of the training data.


Original Abstract Submitted

The network node or the core network may obtain a plurality of datasets for training a ML model, each dataset including a set of metrics collected by a corresponding UE from the at least one UE, and assign at least one data signature associated with a source of each dataset of the plurality of datasets. The network node or the core network may identify a first data signature associated with a corrupted dataset, and filter out at least one dataset associated with the first data signature from the plurality of datasets for training the ML model based on the first data signature being associated with the corrupted dataset.

Retrieved from "http://wikipatents.org/index.php?title=17818315._DATA_SIGNATURES_FOR_ML_SECURITY_simplified_abstract_(QUALCOMM_Incorporated)&oldid=33358"