Abstrait

Data Mining-Based Intrusion Detection Systems

B.Muthulakshmi, Dr.V.Thiagarasu

Intrusion Detection Systems are designed to detect system attacks and it classifies system activities into normal and abnormal form. Data Mining based intrusion detection system model generalizes and detects both known attacks and normal behaviour in order to detect unknown attacks and fails to generalize and detect new attack without known signatures. Intrusion detection system faces three types of issues such as accuracy, efficiency and usability. Intrusion Detection System is used to detect all kinds of new attacks which can be implemented using machine learning techniques with high accuracy. The machine learning techniques are Decision trees, K-nearest neighbour, Ripper rule, Bayesian Network and they are used to analyzed and reducing the false alarm rate. This work compares the accuracy between two intrusion detection systems to examine their Receiver Operating Characteristics (ROC) curves.