Journals / Türkiye Teknoloji ve Uygulamalı Bilimler Dergisi / 2026 / Sayı: 2026

BERT-Based Fake News Detection: An Innovative Approach for Preventing Data Leakage and Methodological Contributions

Bert Tabanlı Sahte Haber Tespiti: Veri Sızıntısını Önleyen Yenilikçi Bir Yaklaşım ve Metodolojik Katkılar

Muhammed Yusuf Kaya(Siirt Üniversitesi Mühendislik Fakültesi)0009-0004-7755-3241
Bashar Alhajahmad(Siirt Üniversitesi Mühendislik Fakültesi)0009-0009-3455-7206
Pages
81–91
Publication Date
2026-07-17
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

In conclusion, it has been verified that BERT models can exhibit superior performance in fake news detection when methodological rigor is maintained; the decisive role of data leakage prevention strategies in model evaluation has been clearly demonstrated.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:The rapid spread of fake news in the digital age has emerged as a critical threat to social trust and democratic processes.
📊
Key Findings & Data:The primary objective is to provide reliable performance evaluation by eliminating methodological errors commonly encountered in the literature.
💡
Conclusion & Impact:In conclusion, it has been verified that BERT models can exhibit superior performance in fake news detection when methodological rigor is maintained; the decisive role of data leakage prevention strategies in model evaluation has been clearly demonstrated.
🔬Design:Modeling & Simulation
📍Scope:1,950 samples
⚙️Variables:sahte haber tespiti · BERT · veri sızıntısı

Abstract

The rapid spread of fake news in the digital age has emerged as a critical threat to social trust and democratic processes. This research presents an innovative BERT-based (Bidirectional Encoder Representations from Transformers) fake news detection approach that systematically addresses the data leakage problem. The primary objective is to provide reliable performance evaluation by eliminating methodological errors commonly encountered in the literature. The ISOT Fake News Dataset, publicly available through the Kaggle platform, was utilized as material. MD5 hash-based duplicate detection, stratified sampling for data splitting, and fine-tuning strategy with BERT-base-uncased model were adopted as methodology. During data cleaning, 6,252 duplicate samples (13.9%) were identified and removed from the dataset, while train-test set overlap was reduced from 1,950 samples to 2. The findings demonstrate that 99.97% test accuracy was achieved on the cleaned dataset, with 99.92% precision and 100% recall obtained for the fake news class. The AUC-ROC score was calculated as 0.9997. In conclusion, it has been verified that BERT models can exhibit superior performance in fake news detection when methodological rigor is maintained; the decisive role of data leakage prevention strategies in model evaluation has been clearly demonstrated.

Özet

Dijital çağda sahte haberlerin hızla yayılması, toplumsal güven ve demokratik süreçler açısından kritik bir tehdit unsuru oluşturmaya başlamıştır. Sunulan araştırma, BERT (Bidirectional Encoder Representations from Transformers) tabanlı yenilikçi bir sahte haber tespit yaklaşımı ortaya koymakta olup, veri sızıntısı (data leakage) problemini sistematik biçimde ele almayı hedeflemektedir. Araştırmanın temel amacı, literatürde sıklıkla karşılaşılan metodolojik hataları gidererek güvenilir bir performans değerlendirmesi gerçekleştirmektir. Materyal olarak herkesin erişimine açık Kaggle platformundan temin edilen ISOT Sahte Haber Veri Seti tercih edilmiştir. Yöntem olarak MD5 hash tabanlı tekrar tespiti, katmanlı örnekleme ile veri bölme ve BERT-base-uncased modeli üzerinde ince ayar (fine-tuning) stratejisi benimsenmiştir. Veri temizliği aşamasında 6.252 tekrarlı örnek (%13.9) belirlenerek veri setinden çıkarılmış, eğitim-test seti örtüşmesi 1.950 örnekten 2'ye indirgenmiştir. Elde edilen bulgular, temizlenmiş veri seti üzerinde %99.97 test doğruluğuna ulaşıldığını, sahte haber sınıfı için %99.92 kesinlik ve %100 duyarlılık değerlerinin sağlandığını ortaya koymaktadır. AUC-ROC skoru 0.9997 olarak hesaplanmıştır. Sonuç olarak, metodolojik titizlik gözetildiğinde BERT modellerinin sahte haber tespitinde üstün performans sergileyebildiği doğrulanmış; veri sızıntısı önleme stratejilerinin model değerlendirmesindeki belirleyici rolü açık biçimde gösterilmiştir.

🔖

Cite This Article

Kaya, M. Y., Alhajahmad, B. (2026). BERT-Based Fake News Detection: An Innovative Approach for Preventing Data Leakage and Methodological Contributions. Türkiye Teknoloji ve Uygulamalı Bilimler Dergisi, Sayı: 2026, 81–91. https://doi.org/10.70562/tubid.1870238

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (19)

  1. Ahmed, H., Traore, I. and Saad, S. (2017). Detection of online fake news using n-gram analysis and machine learning techniques. Proceedings of the International Conference on Intelligent, Secure and Trustworthy Systems, 127-138.
  2. Allcott, H. and Gentzkow, M. (2017). Social media and fake news in the 2016 election. Journal of Economic Perspectives, 31, 211-235.
  3. Devlin, J., Chang, M., Lee, K. and Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 4171-4186.
  4. Graves, L. (2018) Understanding the promise and limits of automated fact-checking. Reuters Institute.
  5. Horne, B. and Adali, S. (2017). This just in: Fake news packs a lot in title, uses simpler, repetitive content in text body, more similar to satire than real news. Proceedings of the International AAAI Conference on Web and Social Media, 11, 759-766.
  6. Horne, B. D., Nørregaard, J. and Adali, S. (2019). Robust fake news detection over time and attack. ACM Transactions on Intelligent Systems and Technology, 11, 1-23.
  7. Kaliyar, R. K., Goswami, A. and Narang, P. (2021). FakeBERT: Fake news detection in social media with a bert-based deep learning approach. Multimed Tools Appl, 80, 11765-11788.
  8. Kapoor, S. and Narayanan, A. (2023). Leakage and the reproducibility crisis in machine-learning-based science. Patterns (N Y), 4, 100804.
  9. Kula, S., Choraś, M. and Kozik, R. (2020). Application of the BERT-based architecture in fake news detection. Proceedings of the 13th International Conference on Computational Intelligence in Security for Information Systems, 239-249.
  10. Lazer, D. M. J., Baum, M. A., Benkler, Y., Berinsky, A. J., Greenhill, K. M., Menczer, F., Metzger, M. J., Nyhan, B., Pennycook, G., Rothschild, D., Schudson, M., Sloman, S. A., Sunstein, C. R., Thorson, E. A., Watts, D. J. and Zittrain, J. L. (2018). The science of fake news. Science, 359, 1094-1096.
  11. Nasir, J. A., Khan, O. S. and Varlamis, I. (2021). Fake news detection: A hybrid CNN-RNN based deep learning approach. International Journal of Information Management Data Insights, 1, 100007.
  12. Rivest, R. (1992) The MD5 message-digest algorithm. Internet Engineering Task Force.
  13. Shu, K., Sliva, A., Wang, S., Tang, J. and Liu, H. (2017). Fake news detection on social media. ACM SIGKDD Explorations Newsletter, 19, 22-36.
  14. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L. ve Gomez, A. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 5998-6008.
  15. Vosoughi, S., Roy, D. and Aral, S. (2018). The spread of true and false news online. Science, 359, 1146-1151.
  16. Wang, W. (2017). Liar, liar pants on fire: A new benchmark dataset for fake news detection. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, 422-426.
  17. Yang, Y., Zheng, L., Zhang, J., Cui, Q., Li, Z. and Yu, P. (2018). TI-CNN: Convolutional neural networks for fake news detection. arXiv preprint arXiv:1806.00749.
  18. Zarocostas, J. (2020). How to fight an infodemic. Lancet, 395, 676.
  19. Zhou, X. Y. and Zafarani, R. (2020). A survey of fake news: Fundamental theories, detection methods, and opportunities. Acm Computing Surveys, 53, 1-40.