International Journal of Engineering and Technology

The Use of internet renders a network packets susceptible to attacks ranging from passive eavesdropping to active impersonation, message replay and message distortion. There is no clear description as to what packets can be considered normal or abnormal. If the intrusions are not detected at the appropriate level, the loss of system may sometimes be unimaginable. Although many intrusion detection system (IDS) methods are used to detect the existing types of attacks within the network infrastructures, reducing false negative and false positive is still a major issue. In this work we present a comparative study between five data mining algorithms to come up finally with the proposition of a hybrid classifier based on Random Forest and Naïve Bayes algorithm. This method provides an effective distinction between different types of intrusions which allows us to customize the treatment given to each type of intrusion. These methods are tested using the KDD'99 database.