Akademik Veri Yönetim Sistemi
GAZI UNIVERSITY JOURNAL OF SCIENCE, cilt.39, sa.1, ss.332-352, 2026 (ESCI, Scopus, TRDizin)
The Smart Grid (SG), a sophisticated electrical network that uses digital technology to monitor and regulate power flow, is susceptible to cyberattacks like eavesdropping, data spoofing, and data falsification. Although there are cryptographic solutions, managing certificate revocation lists (CRLs) is still difficult, and public-key cryptography (PKC) is sometimes unfeasible for inexpensive, power-constrained IoT devices. By taking advantage of hardware flaws in RF devices, Radio Frequency Fingerprint Identification (RFFI) has become a viable non-cryptographic security method. However, for practical implementation, strong deep-learning architectures and efficient deep signal preprocessing are needed. For the first time, we combine XGBoost and BiLSTM in this study to present a hybrid classification framework for RFFI. The accuracy of a Support Vector Classifier (SVC) trained on 15,000 data was 92.6%, whereas the BiLSTM-XGBoost model obtained 97.5% accuracy on 5,000 samples. Furthermore, 97% accuracy was obtained when XGBoost was applied to channel-estimated and equalized wireless data. These findings show how well hybrid deep learning techniques work to strengthen Smart Grid security against online attacks.