EUVIMEDEUVIMED entdecken
Crossref

A Robust Blockchain Network Security Classification Framework

Sultan Ahmad, Mosleh Hmoud Al-Adhaileh, Ghaida Muttshar Abdulsahib

Journal of Eye-Innovation in Security Analysis · 12.08.2026 · 1 (20261) · 29

DOI: 10.68990/jiesa.202613 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Eye-Innovation in Security Analysis, online erschienen am 12.08.2026. Verfasst von Sultan Ahmad und 2 weiteren Autor:innen. Die Arbeit nennt 14 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Distributed networks with blockchain support are increasingly evaluated through joint security, authentication, performance, and energy metrics. Machine-learning benchmarks built from such metrics can still yield deceptively high accuracy if the target is directly constructed from available predictors. A reproducible data audit and classification experiment is carried out on the Blockchain Network Security Dataset. I downloaded the CSV with 1,000 records and 19 columns with no missing values. The target Security_Status is a three-class: At Risk (50.3%), Secure (31.9%), and Moderate (17.8%). This is not the same as the Kaggle description, which describes the target as binary. The depth-four decision audit shows that each label is perfectly reconstructed by four fields: Breach_Status, Anomaly_Score, Decentralized_Authentication_Score and Security_Threat_Level. Thus, two regimes of feature are compared. The full regime comprises all non-identifier predictors; the leakage-audited regime omits the four target-defining fields. Model selection is performed with five-by-five repeated stratified cross-validation on an 80\% training partition, and evaluation is done on a fixed 20\% hold-out. The random forest with all features achieved multiclass macro-F1 scores of 0.9975 + 0.0042 in cross-validation and 1.0000 on the hold-out. After the leakage audit, logistic regression only gets 0.3175 ± 0.0349 macro-F1 and 0.0014 ± 0.0593 Matthews correlation coefficient. For binary Secure-versus-Not-Secure classification, the best audited model has hold-out ROC-AUC 0.5262. These results demonstrate that the reconstruction is being classified by near-perfect classification measures rather than independent threat prediction. The dataset is still useful for data quality education, rule auditing, and exploration of multi-objective metrics, but security benchmarking requires an independently observed outcome, leak-proof predictors, and external validation.

Zitieren

APA

Ahmad, S., Al-Adhaileh, M. H., & Abdulsahib, G. M. (2026). A Robust Blockchain Network Security Classification Framework. Journal of Eye-Innovation in Security Analysis, 1(20261), 29. https://doi.org/10.68990/jiesa.202613

Vancouver

Ahmad S, Al-Adhaileh MH, Abdulsahib GM. A Robust Blockchain Network Security Classification Framework. Journal of Eye-Innovation in Security Analysis. 2026;1(20261):29. doi:10.68990/jiesa.202613

BibTeX
@article{Ahmad2026,
  author = {Ahmad, Sultan and Al-Adhaileh, Mosleh Hmoud and Abdulsahib, Ghaida Muttshar},
  title = {A Robust Blockchain Network Security Classification Framework},
  journal = {Journal of Eye-Innovation in Security Analysis},
  year = {2026},
  volume = {1},
  number = {20261},
  pages = {29},
  doi = {10.68990/jiesa.202613}
}
Kostenlos starten →