IoT Device Identification via Adaptive Feature Set
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing IoT device detection methods face challenges in accurately identifying IoT devices in large-scale environments due to the curse of dimensionality caused by large feature sets and the need for frequent retraining of machine-learning models, as well as limitations in handling new devices and separating training and testing data.
Innovation Solution
The system employs kernel space packet processing to extract variable-length feature sets from packet header information, using a voting classifier and machine-learning models to efficiently classify IoT devices, avoiding the need for manual rule writing and frequent retraining, and leveraging intelligent packet dropping for energy efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large feature sets are used for IoT device detection, then detection accuracy may improve, but computational complexity increases and the curse of dimensionality occurs
Solution Approach 1:
The patent extracts only the most relevant features from packet headers for IoT device detection, rather than using all possible features. This selective extraction reduces the feature set size while maintaining detection accuracy, avoiding the curse of dimensionality and reducing computational complexity.
Solution Approach 2:
The patent applies different feature selection strategies to different packet header fields based on their relevance to IoT device identification. By focusing computational resources on the most informative local features rather than uniformly processing all features, the system achieves accurate detection with reduced overall complexity.
2Adaptability or versatility
If comprehensive feature sets are used to handle diverse IoT devices, then adaptability improves, but model retraining frequency increases
Solution Approach 1:
The patent performs preliminary feature extraction and selection from packet headers before classification, creating a standardized input format that works across diverse IoT devices. This preliminary processing establishes a consistent feature representation that reduces the need for frequent model retraining when new devices are encountered.
Solution Approach 2:
The patent designs a universal feature extraction mechanism that works across multiple IoT device types and protocols. By creating a protocol-agnostic feature set from packet headers, the system achieves broad adaptability without requiring device-specific models, thereby reducing retraining frequency.
3Measurement precision
If more packet data is collected for analysis, then detection accuracy improves, but energy consumption increases
Solution Approach 1:
The patent extracts only essential features from packet headers rather than analyzing complete packet data. This selective extraction maintains detection accuracy by focusing on the most informative fields while significantly reducing the computational energy required for processing.
Solution Approach 2:
The patent applies partial action by analyzing only a subset of packet features that are most relevant to IoT device identification. Rather than processing all packet data, the system performs sufficient analysis on key fields to achieve accurate detection with lower energy consumption.
Data Source
AI summary
Systems and methods for efficient kernel space packet processing and IoT device classification are provided. According to one embodiment, a computer system performs IoT device detection processing. Packet header information is received for multiple packets. Based on the packet header information, multiple Transmission Control Protocol (TCP) or User Datagram Protocol (UDP) flows between a given source device of multiple devices and a given destination device of the multiple devices are identified. For each TCP or UDP flow: a variable-length feature set is created having a size limited by a predetermined or configurable aggregate number of packets sent and received for the TCP or UDP flow; and it is inferred whether the TCP or UDP flow represents an IoT device communication or a non-IoT device communication by applying a machine-learning model to the variable length feature set. The devices are then each classified as either an IoT device or a non-IoT device by aggregating one or more results of the inference processing for each device of the multiple devices with a voting classifier.