Analyzing Knowledge Based Feature Selection to Detect Remote to Local Attacks

Intrusion Detection (ID) is the most significant component in Network Security System as it is responsible to detect several types of attacks. The IDS commonly deals with a large amount of data traffic, which involves irrelevant and redundant features. The feature selection is one of the prominent factors that influence the quality of IDS. We observe that performing feature selection improves the attack detection accuracy as well as the efficiency of the system. In our experiments, we performed manual feature selection, using our domain knowledge with analyzing the nature of the attack. We compare the results of manual feature selection, automatic feature selection and without feature selection for R2L attack. Feature selection finding a subset of features to improve classification accuracy. These features can be used to uniquely identify a specific attack from all the connections. Experimental result on the KDD cup 99 benchmark network intrusion detection dataset demonstrates that the proposed approach achieved high attack detection accuracy. Random Forest is applied on reduced feature set and classification. It is highly accurate classifier. Our proposed work as good as others and time saving for the classification accuracy for R2L attacks.