Transfer Learning Based Intrusion Detection

Abstract

In the past decades, machine learning based intrusion detection systems have been developed. This paper discloses a new aspect of machine learning based intrusion detection systems. The proposed method detects normal and anomaly behaviors in the desired network where there are not any labeled samples as training dataset. That is while a plenty of labeled samples may exist in another network that is different from the desired network. Because of the difference between two networks, their samples produce in different manners. So, direct utilizing of labeled samples of a different network as training samples does not provide acceptable accuracy to detect anomaly behaviors in the desired network. In this paper, we propose a transfer learning based intrusion detection method which transfers knowledge between the networks and eliminates the problem of providing training samples that is a costly procedure. Comparing the experimental results with the results of a basic machine learning method (SVM) and also baseline method(DAMA) shows the effectiveness of the proposed method for transferring knowledge for intrusion detection systems.