Device Classification Rule Export with Differential Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Sharing device classification rules across network deployments is hindered by privacy, security, and resource consumption concerns, particularly in air-gapped networks and bandwidth-constrained environments, where sensitive information and high data volumes are a challenge.
Innovation Solution
A method is introduced where devices replace similar feature vectors with a single vector, apply differential privacy, and send a digest comprising the device classification rule, label, and reduced feature vectors to a cloud service for training a machine learning-based device classifier, ensuring privacy and reducing data volume.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If device classification rules are shared across network deployments, then rule accuracy is improved through cross-validation, but privacy and security concerns worsen due to exposure of sensitive information
Solution Approach 1:
The patent extracts only the essential classification rules and feature vectors from the complete device classification data, separating the useful information needed for cross-validation from the sensitive information that would compromise privacy. This allows sharing minimal necessary data while maintaining rule accuracy improvement.
Solution Approach 2:
The patent introduces an intermediary processing mechanism that transforms raw device classification data into anonymized rule representations before sharing across networks. This intermediary layer prevents direct exposure of sensitive information while still enabling cross-validation benefits.
2Measurement precision
If complete device classification data is exported for training, then machine learning model accuracy is improved, but resource consumption worsens due to high data volumes
Solution Approach 1:
The patent extracts and exports only the essential feature vectors and classification rules needed for effective model training, eliminating redundant and unnecessary data. This extraction process maintains model accuracy while dramatically reducing the volume of data that needs to be transmitted and processed.
Solution Approach 2:
The patent segments the complete device classification dataset into essential feature vectors and classification rules, separating the critical training components from the non-essential data. This segmentation enables selective export of only the necessary elements for model training.
Data Source
AI summary
In one embodiment, a device in a network obtains data indicative of a device classification rule, a device type label associated with the rule, and a set of positive and negative feature vectors used to create the rule. The device replaces similar feature vectors in the set of positive and negative feature vectors with a single feature vector, to form a reduced set of feature vectors. The device applies differential privacy to the reduced set of feature vectors. The device sends a digest to a cloud service. The digest comprises the device classification rule, the device type label, and the reduced set of feature vectors to which differential privacy was applied. The service uses the digest to train a machine learning-based device classifier.