An Alert Data Mining Framework for Network-Based Intrusion Detection System

Abstract

Intrusion detection techniques have been developed to protect computer and network systems against malicious attacks.  However, there are no perfect intrusion detection systems or mechanisms, because it is impossible for the intrusion detection systems to get all the packets in the network system. Current intrusion detection systems cannot fully detect novel attacks or variations of known attacks without generation of a large amount of false alerts. In addition, all the current intrusion detection systems focus on low-level attacks or anomalies. Consequently, the intrusion detection systems usually generate a large amount of alerts. And actual alerts may be mixed with false alerts and unmanageable. As a result, it is difficult for users or intrusion response systems to understand the intrusion behind the alerts and take appropriate actions. The standard format of alert messages is not yet defined. Alerts from heterogeneous sensors have different types although they are actually same. Also false alarms and frequent alarms can be used as Denial of Service attack as alarm messages by themselves and cause alert flooding. So we need to minimize false alarm rate and prevent alert flooding through analyzing and merging of alarm data. In this paper, we propose a data mining framework for the management of alerts in order to improve the performance of the intrusion detection systems. The proposed alert data mining framework performs alert correlation analysis by using mining tasks such as axis-based association rule, axis-based frequent episodes and order-based clustering. It also provides the capability of classifying false alarms in order to reduce false alarms from intrusion detection system. The final rules that were generated by alert data mining framework can be used to the real time response of the intrusion detection system and to the reduction of the volume of alerts.