Smart Intrusion Detection System for DMZ

Prediction of network attacks and machine understandable security vulnerabilities are complex tasks for current available Intrusion Detection System [IDS]. IDS software is important for an enterprise network. It logs security information occurred in the network. In addition, IDSs are useful in recognizing malicious hack attempts, and protecting it without the need for change to client‟s software. Several researches in the field of machine learning have been applied to make these IDSs better a d smarter. In our work, we propose approach for making IDSs more analytical, using semantic technology. We made a useful semantic connection between IDSs and National Vulnerability Databases [NVDs], to make the system semantically analyzed each attack logged, so it can perform prediction about incoming attacks or services that might be in danger. We built our ontology skeleton based on standard network security. Furthermore, we added useful classes and relations that are specific for DMZ network services. In addition, we made an option to mallow the user to update the ontology skeleton automatically according to the network needs. Our work is evaluated and validated using four different methods: we presented a prototype that works over the web. Also, we applied KDDCup99 dataset to the prototype. Furthermore,we modeled our system using queuing model, and simulated it using Anylogic simulator. Validating the system using KDDCup99 benchmark shows good results law false positive attacks prediction. Modeling the system in a queuing model allows us to predict the behavior of the system in a multi-users system for heavy network traffic