IoT Device Identification Using Directional Packet Length Sequences and 1D-CNN.

Area of Science:

• Computer Science
• Cybersecurity
• Network Engineering

Background:

• The proliferation of Internet of Things (IoT) devices presents significant security challenges.
• Effective IoT device identification is crucial for securing networks and managing devices.
• Current passive fingerprinting methods often overlook crucial packet sequence information like direction and length.

Purpose of the Study:

• To propose a novel IoT device identification method utilizing directional packet length sequences.
• To leverage deep convolutional neural networks (CNNs) for enhanced feature extraction from network traffic.
• To improve the accuracy and effectiveness of IoT device recognition.

Main Methods:

• Constructing device fingerprints based on directional packet length sequences in network flows.
• Employing a deep convolutional neural network (CNN) to analyze these fingerprints.
• Extracting deep features using convolutional layers for classification.

Main Results:

• The proposed method achieved over 99% accuracy, recall, precision, and F1-score in device identification.
• Experimental results demonstrate high effectiveness in recognizing diverse IoT device identities.
• The approach proved superior to traditional machine learning methods in feature representation and classification.

Conclusions:

• Directional packet length sequences offer an intuitive and effective feature representation for IoT device identification.
• Deep convolutional neural networks are well-suited for extracting complex patterns from these sequences.
• The developed method provides a robust solution for enhancing IoT security through accurate device recognition.