Features selection for intrusion detection system using hybridize PSO-SVM

An Intrusion Detection System is software or application which is used to detect thread, malicious activities and the unauthorized access to the computer system and warn the administrators by generating alarms. Features selection process can be considered a problem of global combinatorial optimization in machine learning. Genetic algorithm GA had been adopted to perform features selection method; however, this method could not deliver an acceptable detection rate, lower accuracy, and higher false alarm rates. Hybridize Particle Swarm Optimization (PSO) as a searching algorithm and support vector machine (SVM) as a classifier had been implemented to cope with this problem. The results reveal that the proposed hybrid algorithm is capable of achieving classification accuracy values of (95.82 % and 97.68 %), detection rates values of (95.8 % and 99.3 %) and false alarm rates values of (0.083 % and 0.045 %) on both KDD CUP 99 and NSL KDD. Electing the best set of features will help to improve the classifier predictions in terms of the normal and abnormal pattern. The simulation will be carried on WEKA tool, which allows us to call some data mining methods under JAVA environment. The proposed model will be tested and evaluated on both NSL-KDD and KDD-CUP 99 using several performance metrics.

Preview Text FSKTM 2017 22 IR.pdf Download (1MB) | Preview

Actions (login required)

View Item