Abstract:

Abstract By allowing attackers to take control of a significant amount of infected devices for illegal purposes, botnets pose severe challenges to network security. Due to the dynamic nature of botnet infrastructures and the advanced tactics used by attackers, identifying and preventing attacks by botnets is a complex undertaking. In this study, we present a novel machine learning (ML)-based black hole optimized random forest (BHO-RF) for botnet detection. The BHO method enhances the performance of the RF classifier by modifying the hyperparameters, boosting its ability to recognize botnet traffic. It is inspired by the behaviour of black holes in space. We have extensive tests utilizing the CTU-13 dataset to assess the efficacy of our suggested framework. The outcomes show that our multi-layered strategy outperforms conventional techniques, delivering higher accuracy, f1-score, precision, and recall in botnet identification. The approach also demonstrates robustness over noise and changes in transmission characteristics.

Keywords: