Correlation-Based Feature Selection on Botnet Activity Detection Using Kendall Correlation

Botnets are a dangerous threat to computer networks that uses malicious code to infect computer networks. Thus, the right system security model is needed to detect botnet attack activities accurately. Several previous studies have introduced a botnet detection model using mining-based, but it requires the correct approach to obtain the optimal performance. This paper proposes a botnet detection model by improving feature selection using correlation-based analysis. The aim is to improve accuracy detection by analyzing features with solid correlations that can be used for machine learning classification models. The proposed model consists of 4 main parts: data splitting pre-processing, classification process, and evaluation. The experiment used public datasets, namely CTU-13 datasets containing botnet activity. The experiment shows that the model can detect botnet activity with a detection accuracy of 99.7218%, precision of 99.1691 %, and recall of 96.6533 %. The proposed model can improve the existing botnet detection system model.