Anomaly Detection in IoT Based Satellite Networks: NidaDeepMix

Abstract

IOT based satellite networks are one of the modern cyber attack topics. This technology has important application areas such as data collection, monitoring and control without the need for close access. Especially the increasing use of IOT devices and their recent integration with satellite networks have made these devices the target of attacks. The fact that IOT devices have more than one type and require low processing power makes them vulnerable to attacks. The use of IOT devices together with satellite networks increases the complexity of this situation and the size of cyber attacks. This situation has made it necessary to increase the studies on preventing and detecting cyber attacks on IOT based networks. For this purpose, in this article, we propose a new deep learning architecture (NidaDeepMix) that provides high accuracy in order to detect cyber attacks on IOT based satellite networks. The designed layer structure and parameters of the NidaDeepMix architecture are adjusted to effectively cope with complex and difficult situations. The NidaDeepMix architecture has been tested on two separate comprehensive datasets, CSE-CIC-IDS-2018 and BCCC-CIRA-CIC-DoHBrw-2020. As a result of the training, a serious accuracy rate of %99.99 was achieved for the CSE-CIC-IDS-2018 dataset and %99.98 for the BCCC-CIRA-CIC-DoHBrw-2020 dataset. Considering these high accuracy rates, it has been demonstrated that the proposed architecture is quite effective in classifying attacks. These rates obtained on different datasets reveal the generalization success of the model. At the same time the model has also addressed the issue of cyber attacks on IOT based satellite networks with an innovative approach. In this context, a new and effective architecture has been provided to the literature for detecting attacks on IOT based satellite networks. It is envisaged that the proposed method NidaDeepMix will be an important reference model in important issues such as cyber attacks and anomaly detection in the future. © 2025, Modern Education and Computer Science Press. All rights reserved.