Detecting cyber threats in iot based satellite networks

Satellite based IoT networks are of paramount importance in cybersecurity due to their use in many critical services such as communication and defense. The aim of this thesis is to develop an integrated analysis approach that can detect cyber threats occurring in IoT based satellite networks early and with high accuracy. This study tests the hypotheses that hybrid deep learning models will yield more successful results compared to traditional methods by learning both temporal and structural patterns together, and that multi layered architectures will improve performance in distinguishing different types of attacks. The research consists of network traffic logs obtained from different cybersecurity datasets representing IoT and satellite network environments. These datasets were selected to include both normal and various attack types and are reliable datasets frequently used in current studies. A multi stage method consisting of data preprocessing, feature extraction and classification stages was applied in the study. Anomaly detection was performed using hybrid deep learning models combining methods such as CNN, LSTM and attention mechanisms and the reliability of the model was evaluated with various performance metrics. The findings demonstrate that the proposed models can distinguish between attack and normal traffic with high accuracy rates and yield consistent results across different datasets. Hybrid structures were found to be more effective than single models in detecting complex attack patterns. This thesis presents multiple scalable and highly accurate cyber threat detection approaches for IoT based satellite networks.