Use of Machine Learning Algorithms with SIEM for Attack Prediction

Abstract

In the recent years, organizations face the ever growing challenge of providing security in the network infrastructure. An intrusion detection system is essentially a spruced up, intelligent variant of a firewall which does deep packet analysis which generate alerts but cannot predict multistep attacks. In this work, we propose an intrusion prediction system (IPS) with the extension of a commercial SIEM framework, namely open source security information management (OSSIM), to perform the event analysis and to predict future probable multistep attacks before they pose a serious security risk. Security information and event management (SIEM) framework affirms network protection by the correlation and management of network log files. Data mining techniques are used for processing of all normalized data from OSSIM and also for classification.