Intelligent intrusion detection system using clustered self organized map

The impact of information security breaching becomes bigger and complicated to ignore every day. New and more sophisticated attacks are emerging and developed; requiring the information systems and networks be protected in a highly flexible and accurate manner. Intrusion Detection Systems (IDS) are considered one of the basic building blocks of the protection wall against these intrusive activities through detecting it before it hits the network systems. Artificial neural networks have been used successfully for addressing the high accuracy and precision demands of intrusion detection systems. In this paper, we built an intelligent intrusion detection system using clustered version of Self-Organized Map (SOM) network. The proposed system consists of two subsequent stages: first, SOM network was built, then a hierarchical agglomerative clustering using k-means was applied on SOM neurons. The proposed work in this research paper addresses the issues of sensitivity and time consumption for each connection record processing. The proposed system was demonstrated using NSL-KDD benchmark dataset, where it has achieved superior sensitivity reached up to 96.66 % in less than 0.08 milliseconds per connection record.