Network traffic classification based on deep learning

Date of Publication

8-3-2023

Document Type

Master's Thesis

Degree Name

Master of Science in Information Technology

Subject Categories

OS and Networks

College

College of Computer Studies

Department/Unit

Information Technology

Thesis Advisor

Danny Cheng

Defense Panel Chair

Fritz Kevin Flores

Defense Panel Member

Gregory Cu
Danny Cheng

Abstract/Summary

With the current era of rapid network expansion, network traffic is increasing day by day, posing challenges to network management and applications. Network traffic classification is an important prerequisite for network operation management, traffic intrusion detection, and user behavior analysis. At present, most network traffic classification technologies are based on traditional machine learning methods. The classification accuracy is highly dependent on the design of traffic feature sets, and the selection of effective feature sets requires rich experience in feature engineering. In recent years, with the further development of deep learning, it has been widely used in the fields of computer vision, natural language processing and speech recognition. However, the deep learning framework has strict requirements on the format and size of the input data, so the process needs to be preprocessed first. At present, most of the process preprocessing processes have defects such as redundant input data and excessive scale, which eventually lead to long training time of deep learning models and excessive model calculations. This paper mainly studies the network traffic classification method based on RNN and CNN models. The main work is as follows: This paper proposes to add a sequence-sensitive RNN to pre-train the traffic before CNN classification, and use the trained model to pre-process the network traffic and generate grayscale images or other formatting as the next step. Input to CNN. In this way, by adding RNN, it can make up for the problem that CNN cannot fully learn the traffic data structure and timing characteristics. Overall, the proposed approach will involve using RNNs to extract features from sequential network traffic data, generating grayscale images to represent temporal dynamics, and then using CNNs to classify the data based on the extracted features. This approach has the potential to improve the accuracy of network traffic classification and can be applied to a wide range of network security applications.

Keywords: network traffic classification, network data preprocessing, deep learning, RNN, CNN

Abstract Format

html

Language

English

Format

Electronic

Physical Description

47 leaves

Keywords

Computer networks; Deep learning (Machine learning)

Upload Full Text

wf_yes

Embargo Period

8-12-2023

This document is currently not available here.

Share

COinS