查看文章

The number of Internet-of-Things (IoT) devices has significantly expanded as a result of the growing reliance on the Internet and the associated rise in connectivity demand. According to recent research, the increasing adoption of IoT devices has in turn increased network threats since there are more possible attack surfaces. This demonstrates how IoT networks and devices are becoming more vulnerable and susceptible. As a result, in such situations, proper, effective, and efficient attack detection and mitigation approaches are required. IoT devices' security flaws make it simple for attackers to take advantage of them and incorporate them into a botnet. Once hundreds of thousands of IoT devices have been hijacked and joined a botnet, attackers utilize this botnet to perform sophisticated distributed denial of service (DDoS) assaults that bring down the target websites or services and prevent them from responding to legitimate users. Numerous botnet detection approaches have been put out thus far, but their effectiveness is constrained by the dataset on which they are trained. Due of the variety of attack methods, the features used to train a machine learning (ML) model on a botnet dataset do not perform well on other datasets. Therefore, irrespective of the underlying dataset, we suggest an uniform characteristics set in this work to better detect botnet assaults. When the trained ML models were evaluated over three distinct datasets for botnet attacks, the suggested feature set demonstrated superior results for identifying botnet attacks.