.png){kind=icon}
.png){kind=icon}
Return
SARA: A Lightweight, Unknown-Oriented, SemiSupervised Network Traffic Classification Approach Based on Ensemble Multiple Sub-classifiers Learning
.png){kind=icon}
PRE
AI
DOI：10.22967/HCIS.2024.14.038.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Machine learning (ML) technology is critical to flow statistical features for network traffic classification. Unfortunately, much work has been focused on using ML techniques to solve the problems related to the classification of network traffic, such as a severe degradation in classification performance and the difficulty of identifying unknown flows in the real world. Such problems are mainly caused by flow features' high dimensionality and redundancy, the imbalanced number of traffic classes, concept drift due to Internet traffic, and over reliance on labeled data. To overcome these problems, this paper proposes a novel approach based on the idea of semi -supervised learning for network traffic classification, called SARA (feature Selection, deep Auto -coder, Redundancy Analysis), which introduces deep learning, efficient feature processing methods and a heuristic traffic classification model in order to achieve an outstanding overall optimization. This paper shows analytically that SARA can solve the unknown flow identification problem in most situations. Based on real traffic traces, the experimental results show that the proposed approach can efficiently reduce a feature space's dimensions and deal with the multi -class imbalance and concept drift problems in terms of ML. Furthermore, it can improve the classification performance and have a beneficial classification effect on unknown flows.
Keywords:
Network Traffic Classification
Semi-supervised Learning
Feature Selection
Deep Auto-coder
Clustering Analysis
Journal
IF:
3
Papers:
555
Citations:
1.4K