An Enhanced Approach For Botnet Intrusion Detection System Based on Machine-Learning Model

In this cyber era, the speed of technological advancements is likely linear to the constant threat to information security, one of which is the botnet attack. Recently, botnet advancement has been developing toward nowhere, where it evolves in size and sophistication. Numerous botnet Intrusion Detection Systems (IDS) have been developed, especially one based on network-flow detection. However, the nature of the recent botnet, equipped with advanced code updates, makes it difficult for the IDS to keep track of the botnet attack. Thus, abnormal botnet detection methods are better because they can mark a new or uncharted bot flow. This research proposes a new approach that can detect botnet types. A generic model that consists of network-flow data pre-processing and feature selection is introduced. It is then installed to the pre-known machine learning classification techniques. The model hence uses the CTU-13 dataset to measure the performance of botnet detection. Experimental results show that this proposed work works as intended in botnet detection, and the decision tree algorithm produced the best average detection accuracy of 99.02%.