A New Approach Feature Selection for Intrusion Detection System Using Correlation Analysis

Threats and attacks on computer networks need to be handled properly in the cyber era. Malicious activities can harm the availability of system resources to the company's financial losses. Anticipating the malicious effects of attacks can be done by developing an attack detection model known as an Intrusion Detection System (IDS). Several previous studies have developed attack detection models with various detection optimization methods, for example, by optimizing the feature selection process. However, it has not shown a correlation analysis between features to get a strong correlation that can affect the improvement of detection performance. This paper proposes an attack detection model by developing a correlation-based feature selection technique that adopts the Pearson correlation algorithm. Correlation analysis was developed by measuring the correlation threshold to get a strong correlation between features, and its selected for classification modeling using Random Forest. The result of feature selection is obtaining 35 features, and the detection accuracy is 99.8136%, precision is 99.9687%, and recall is 99.7039%. The proposed model can be used to develop the existing intrusion detection model.