MiraiBotGuard: Federated Learning for Intelligent Defense Against Mirai Threats

With the advancement of technology information has taken the shape of digital information that can belong to industries, businesses and other organizations. The information can be breached with various types of attacks and Internet of Things devices are widely used as equipment to target the victim machine which is storing the critical information. Artificial Intelligence-based methods have become frequent in providing cybersecurity solutions to the masses. The issue that arises with artificial intelligence methods is the need to share internal network logs with third-party or learning models. In this research work, the Mirai attack detection methodology is proposed using the federated learning approach with the deep neural network. The proposed model is evaluated using the CIC-IoT 2023 dataset by filtering network traffic related to the Mirai attack dataset. Further, the approach is evaluated for two scenarios covariate shift and concept drift. The local and global learning is enhanced using the federated averaging technique. The evaluation results are compared and studied based on metrics of accuracy, loss, precision and recall. The robustness and reliability of the proposed approach are proved with consistently improved results for state metrics and also the performance for concept drift outperforms the covariate shift.