Akademik Veri Yönetim Sistemi
5th International Conference on Modern and Advanced Research - ICMAR 2026 , Konya, Türkiye, 7 - 08 Mayıs 2026, ss.211-216, (Tam Metin Bildiri)
This study addresses the simulation of attacks threatening the physical layer security of 6G wireless networks, the generation of labeled datasets, and machine learning-based attack detection. As part of this study, a multi-layer 6G topology featuring a Satellite-Air-Ground-Sea (SAGS) architecture was modeled using the NS-3 network simulator. A total of 12 attack types—including Jamming, Eavesdropping, Spoofing, DoS, Pilot Contamination, RIS Manipulation, MITM, Replay, Wormhole, Sinkhole, Blackhole, and DDoS—were simulated on this topology. A labeled dataset consisting of 33,953 samples and containing 25 features was generated from the simulations. The Random Forest, Decision Tree, and Support Vector Machine algorithms were compared on the generated dataset. According to the results, the Random Forest model demonstrated the most balanced performance with approximately 76% accuracy, 0.763 precision, 0.760 recall, and an F1 score of 0.76. The findings indicate that it is possible to model 6G-specific threats alongside classical network attacks within the same dataset and that machine learning-based approaches are applicable for real-time attack detection.