A Scalable XGB-RF Approach For Multi-Class Iot Botnet Detection
DOI:
https://doi.org/10.64252/ptwh0v88Keywords:
IoT network security, botnet intrusion detection, feature selection, XGBoost, Random Forest (RF), Mirai, DDos.Abstract
IoT devices have changed very quickly in the last few years, which has made them more useful than ever for making life easier. But there are still a lot of security holes in IoT devices. This is mostly because most of them don't have the memory or processing power needed to support strong security features. Because of this, all kinds of cyberattacks can get to IoT devices. A single breach of a network system or device can put data security and privacy at risk. But you can use machine learning algorithms to find attacks on IoT devices. This work proposes a multimodal machine learning model called XGB-RF that can be used to find intrusion attacks. The N-BaIoT dataset, which includes harmful botnet attacks, was used with the proposed hybrid framework. The random forest (RF) algorithm was used to choose attributes, and the eXtreme Gradient Boosting (XGB) classifier was used to sort different kinds of attacks in IoT ecosystems. We use different performance measures to see how well the proposed XGB-RF model works. The results show that the model finds 99.94% of the attacks correctly. The proposed model does better on all the measures than the best algorithms that are currently available. The proposed method has a lot of potential to make IoT systems safer because it can find botnet attacks very well.