Botnet-based IoT Network Attacks Identification using LSTM

Internet of Things (IoT) networks are growing significantly as a modern communication technology that changes the world tremendously, and the vulnerabilities of IoT have brought many cyber attacks. Botnet attacks are considered a major security threat to IoT networks, with the potential to cause significant damage and disruption. Classifying botnet traffic in IoT networks is vital for maintaining network security. Deep learning (DL) shows its dominance in cyber-security, so this paper presents a Long Short-Term Memory (LSTM) model to classify botnet-based IoT networks. The proposed model is evaluated using two benchmark network intrusion datasets, i.e., CICIDS-2017 and N-BaIoT dataset. The experiment carried out on the CICIDS-2017 gives the classification accuracy as 99.76% and 99.38% on Wednesday data and the whole dataset of the CICIDS-2017 dataset, respectively. Moreover, we applied a statistical analysis measure called Matthew coefficient correlation (MCC) to evaluate our model and found the MCC scores as 99.52% and 98.76% on Wednesday data and the whole dataset, respectively. The proposed model also performs better than the compared models while using the N-BaIoT dataset, achieving a 99.98% accuracy score as well as a 99.95% MCC score.