Detecting Transport and Application Layer DDos Attacks in IoT Devices with Machine and Deep Learning

Many aspects of daily life are being revolutionized by the Internet of Things, from electronic houses to industrial settings. However, considering the vast range of manufacturers involved, concerns over security standards have been raised by the expanding number of devices linked to the Internet. Denial of Service attacks are a frequent and significant security concern that may affect IoT networks. It's interesting to note that the latter three months of 2021 had a 31% rise in sophisticated DDoS targeted attacks over the same period in 2020. This study builds a creative intrusion detection system using deep learning models and machine learning on the Bot-IoT dataset to solve issues about class imbalance. Three distinct feature sets are employed for binary and multiclass classifications to assess how the timestamp affects predictions. Through the use of this technique, the inherent feature dependencies of the Argus flow info generator are minimized, resulting in an average accuracy rate above 99%. The recommended models perform better at identifying denial-of-service attacks than the state-of-the-art, according to comprehensive testing that includes temporal performance analysis. Notably, decision tree and multi-layer perceptron models are the most effective methods for detecting DDoS and DoS attacks on IoT networks