Abstract

Using the network connection records for intrusion detection, not only the data dimension is high, but also the difference between different classes is big in quantity. Visual intrusion detection method based on weighted non-negative matrix factorization can not only reduce dimension, but also classify imbalanced data set effectively. After network connection records are mapped from high dimensional space to low dimensional visual space by weighted non-negative matrix factorization, their characteristics are shown in low dimensional visual space by scatter dots. Visual intrusion detection is realized by observing the located area of scatter dots corresponding to the detected network connection records. The effectiveness of this method has been demonstrated by experiments.

Paper Details