Loading...

Network Traffic Classification using Test Input Generation and Time-Related Feature Generation

Arabi Jashoughani, Mohammad Mahdi | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57574 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Fazli, Mohammad Amin
  7. Abstract:
  8. Network traffic classification plays an important role in network monitoring and management. With the continuous development of network technology, traditional traffic classification methods face increasing limitations in accuracy, especially when dealing with encrypted traffic. Fortunately, deep neural networks offer an effective method for traffic classification due to their ability to learn the intrinsic features of data. In this study, we use the public ISCX network traffic dataset for our evaluations. We examine two general approaches. The first approach involves converting traffic into MNIST images and classifying classes using convolutional neural networks (CNNs), generating test inputs, and training the model with the generated inputs. The second approach uses general packet information such as time and packet volume, generating time-related features, and classifying using long short-term memory (LSTM). Both approaches use different training and evaluation datasets. Various algorithms are evaluated for each approach. The primary goal of this research is to develop a comprehensive model suitable for unknown network datasets. Simulation results show that the CrossPoint algorithm achieved 87% accuracy with the test data generation and model training approach, while the DeepFlow algorithm achieved 88.5% accuracy using the time-related feature generation approach with extended training time
  9. Keywords:
  10. Network Traffic ; Traffic Classification ; Deep Neural Networks ; Long Short Term Memory (LSTM) ; Convolutional Neural Network ; Network Management ; Adversarial Testing

 Digital Object List

 Bookmark

No TOC