Robot Communication: Network Traffic Classification Based on Deep Neural Network.

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• Robot communication security is increasingly critical due to widespread robot adoption.
• Existing traffic classification methods struggle with encrypted data and require manual feature engineering.
• There is a need for automated, effective encrypted traffic classification methods.

Purpose of the Study:

• To propose an automated traffic classification framework for encrypted robot communication.
• To develop a method that eliminates the need for manual feature extraction.
• To enhance the accuracy of encrypted traffic recognition.

Main Methods:

• A novel traffic classification framework utilizing a capsule neural network (CNN) with multilayer neural networks.
• Implementation of capsule vectors for enhanced data stream characteristic learning.
• A hybrid classification network combining CNN and long short-term memory (LSTM) networks to capture temporal and spatial traffic features.

Main Results:

• The proposed capsule neural network framework effectively classifies encrypted network traffic.
• The model automatically learns data stream characteristics without manual feature extraction.
• The combined CNN-LSTM structure successfully learns network traffic time and space characteristics.
• An 8% increase in recognition accuracy was achieved compared to previous methods.

Conclusions:

• The developed framework provides an effective solution for classifying encrypted robot communication traffic.
• Automated feature learning significantly improves the efficiency and accuracy of traffic classification.
• The capsule neural network and hybrid CNN-LSTM approaches represent a significant advancement in network security for robotic systems.