PAHDF: A Privacy-Aware Hybrid Detection Framework with Class-Aware Weighted Stacking Ensemble (CAWSE) for Fake Instagram Account Detection

Sura Jasim Mohammed, Department of Software, College of Information Technology, University of Babylon, Babylon, IraqFollow
Safa Saad Abbas, Department of Software, College of Information Technology, University of Babylon, Babylon, IraqFollow
Suhad Hatem Jihad, Computer Center, University of Babylon, Babylon, IraqFollow

ORCID

Sura Jasim Mohammed: https://orcid.org/0000-0002-8376-2822

Safa Saad Abbas: https://orcid.org/0000-0002-3119-4317

Suhad Hatem Jihad: https://orcid.org/0009-0008-3268-2970

Article Type

Original Study

Abstract

The rapid growth of social media platforms has intensified concerns regarding online privacy, data security, and fraudulent activities that driven by fake accounts. This paper proposes a Privacy-Aware Hybrid Detection Framework (PAHDF) to detect Instagram fake account that integrates privacy preservation with high-performance machine learning. Unlike existing approaches that treat privacy and detection as separated objectives, therefore, the proposed framework jointly addresses both objectives by relying exclusively on publicly available, low-sensitivity profile metadata. PAHDF combines a deep learning model for latent feature representation with a Random Forest classifier for behavioural pattern learning through a Class-Aware Weighted Stacking Ensemble (CAWSE), where adaptive class-specific weights improve minority-class detection. The framework further incorporates a hybrid class-imbalance strategy based on undersampling, oversampling, and cost-sensitive learning to improve classification robustness. Experiments were conducted on a manually annotated dataset comprising 10,000 Instagram accounts (6,868 legitimate and 3,132 fraudulent). The proposed framework achieved approximately 94.1% validation accuracy, 92.5% balanced accuracy, 0.926 F1-score, 0.961 ROC-AUC, and a Matthews Correlation Coefficient (MCC) of 0.789, demonstrating reliable performance under class-imbalanced conditions. An ablation study further confirmed that the proposed UOC + CAWSE strategy consistently outperformed individual imbalance-handling techniques. To improve model interpretability, feature importance was analysed using both Random Forest Mean Decrease in Impurity (MDI) and SHAP, with a Spearman rank correlation of ρ = 0.993 (p < 0.001) confirming strong agreement between the two explainability methods. Comparative evaluation against fourteen state-of-the-art approaches demonstrated the effectiveness of PAHDF while preserving user privacy. These findings indicate that accurate, interpretable, and privacy-preserving fake account detection can be achieved using only publicly accessible account metadata.

Keywords

Fake account detection, Privacy-aware machine learning, Stacking ensemble, Deep learning, Class imbalance, Instagram security, Social media protection

How to Cite This Article

Mohammed, Sura Jasim; Abbas, Safa Saad; and Jihad, Suhad Hatem (2026) "PAHDF: A Privacy-Aware Hybrid Detection Framework with Class-Aware Weighted Stacking Ensemble (CAWSE) for Fake Instagram Account Detection," Journal of Intelligent Informatics, Networking, and Cybersecurity: Vol. 2 : Iss. 2 , Article 11.
Available at: https://doi.org/10.65445/3106-1192.1022