| Summary: | 碩士 === 逢甲大學 === 資訊工程學系 === 103 === Network traffic classification has been an indispensable technique for network management and network security. Especially with the rapid evolution of the Internet in recent years, there are zero-day traffic flows that appear on Internet almost every day. Zero-day traffic flows are the flows which belong to unknown applications and might have serious impact on the network Quality of Service (QoS). However, it requires a lot of time to train classifiers to improve classification performance with supervised machine learning classification methods. Unsupervised machine learning clustering methods have some problems, such as an uncertain number of clusters, cluster correctness and computational complexity. Due to these limitations, we propose an aggregation clustering algorithm by grouping similar flows into clusters. The proposed method is able to evaluate the correlation between flows and reduce the amount of noise. Experimental results show that the proposed method can determine the number of clusters and still have decent classification accuracy.
|
|---|
