Securing IoT-Edge Networks: Federated Deep Learning for Botnet Detection

Governments throughout the world are encouraging the use of “smart city” technologies to improve urban residents' day-to-day experiences. In order to improve services like healthcare, electricity distribution, water purification, traffic management, etc., smart cities implement internet-connected technologies. The proliferation of connected gadgets has led to an increase in botnet assaults based on the IoT. The term IoT is used to describe a system of computers that are linked together that may be used for a wide variety of tasks, from environmental monitoring to on-demand power switching and beyond. Many Internets of Things gadgets are inherently disparate, update at irregular intervals, and hide in plain sight on a private or company network. The safety and confidentiality issues surrounding the Internet of Things that need to be addressed in both academic and practical settings. This research study proposes a federated-based solution to botnet attack detection utilizing on-device decentralized traffic data and a deep learning (DL) model. The proposed federated method addresses privacy concerns by preventing data from leaving the network's edge on the device. Instead, the edge layer is used to do the DL calculation, which has the extra benefit of being closer to the source of the data. Many tests are run on newly made public test data sets for deep learning models. Additionally, the sets of data are presented for examination and understanding. The recommended DL model achieved better results than the ML models. Finally, this research shows that the suggested model can detect anomalies with a precision of up to 98%.