Active Learning for IoT Device Type Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As the Internet of Things (IoT) expands, existing device classification systems face challenges in accurately identifying new device types due to the proliferation of diverse IoT devices, leading to a high percentage of devices being categorized as 'UNKNOWN', which is difficult to manage and scale, especially with the increasing number of endpoint devices.
Innovation Solution
The approach involves clustering traffic feature vectors for endpoints in a network, selecting clusters for labeling based on contextual data, and using a machine learning-based device type classifier trained with minimal user input, allowing for adaptive selection of clusters for efficient labeling and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional device classification systems are used to identify IoT devices, then the system can handle known device types, but the percentage of devices categorized as UNKNOWN increases with the proliferation of new device types
Solution Approach 1:
The system performs unsupervised clustering on traffic feature vectors before supervised classification, pre-grouping unknown devices into clusters based on behavioral similarities. This preliminary action enables the system to handle new device types by grouping them with similar unknown devices, preventing them from being classified as UNKNOWN while maintaining accurate classification for known devices.
2Measurement precision
If manual labeling of all unknown devices is performed, then device type accuracy improves, but the time and resources required for labeling increase significantly
Solution Approach 1:
The system applies supervised classification only to representative samples from each cluster (e.g., one or a few devices per cluster) rather than labeling every unknown device individually. This partial action approach maintains high labeling efficiency while achieving accurate device type identification, as the labeled samples are used to train classifiers that can then automatically classify similar devices in the same cluster.
3Measurement precision
If more devices are labeled for training data, then the machine learning classifier becomes more accurate, but the complexity and cost of data collection increase
Solution Approach 1:
The system segments the large dataset of unknown devices into multiple clusters based on traffic feature similarities. By dividing the data collection effort into cluster-specific samples rather than requiring comprehensive labeling of all devices, the system reduces data collection complexity while maintaining classifier accuracy through targeted sampling from each segment.
Data Source
AI summary
In one embodiment, a device clusters traffic feature vectors for a plurality of endpoints in a network into a set of clusters. Each traffic feature vector comprises traffic telemetry data captured for one of the endpoints. The device selects one of the clusters for labeling, based in part on contextual data associated with the clusters that was not used to form the clusters. The device obtains a device type label for the selected cluster by providing data regarding the selected cluster and the contextual data associated with that cluster to a user interface. The device provides the device type label and the traffic feature vectors associated with the selected cluster for training a machine learning-based device type classifier.