Performance analysis of an Intrusion Detection System using Panjab University Intrusion DataSet

Abstract

Intrusion Detection System (IDS) can be used to detect malware by its network activities or behavioral profiles. Common challenges for IDS are large amount of data to process, low detection rate and high rate of false alarms. Online Sequential Extreme Learning Machine (OS-ELM) based IDS with network traffic profiling is tested on Panjab University - Intrusion DataSet (PU-IDataSet). This IDS is known as alpha-FST-Beta IDS. The training connections are first categorized on the basis of protocol and service features. This categorization is named as alpha profiling. It increases the scalability and reduces the time complexity of IDS. Large feature set of network traffic dataset is reduced using ensemble of three feature selection techniques. Beta profiling is used to reduce the size of training dataset. Various parameters like accuracy, true positive rate, false positive rate, true negative rate, false negative rate, precision, F1-score and detection time is used to evaluate the performance. The results obtained encourage the integration of this system in intrusion detection models.