Cold-Start Fake Account Detection on Instagram Using Deep Learning: A Systematic Review of Methods and Gaps
DOI:
https://doi.org/10.47941/ijce.3400Keywords:
Cold-start Detection, Fake Instagram Accounts, Deep Learning, Social Media Fraud, Account Impersonation, GANs, Autoencoders, User Behavior AnalysisAbstract
Purpose: The primary objective of this systematic review is to evaluate the effectiveness of Deep Learning (DL) architectures in detecting "cold-start" fake accounts on Instagram newly created profiles that lack sufficient historical data for traditional detection.
Methodology: The methodology focused on five core DL frameworks Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, Generative Adversarial Networks (GANs), and Autoencoders evaluating their ability to process non-textual features, metadata, and early-stage behavioral patterns.
Findings: The findings reveal that hybrid models, specifically those combining GANs for data augmentation with LSTMs for sequence analysis, achieve the highest detection accuracy of up to 96.4% for cold-start profiles. However, a significant transparency-accuracy trade-off persists, as ensemble methods often lack the interpretability required for platform-wide implementation and struggle to distinguish between "quiet" legitimate users and sophisticated "human-impersonating" bots during the critical first 48 hours of account activity.
Unique Contribution to Theory, Policy and Practice: This study contributes to theory by introducing an integrated framework for "digital identity evolution" that moves beyond static feature analysis toward dynamic, behavior-based detection. In practice, it provides platform developers with a technical roadmap for implementing hybrid-optimization models, such as combining DL with bio-inspired algorithms like GWO and PSO, to reduce false-positive rates. Finally, for policy, the research offers evidence-based recommendations for regulatory frameworks regarding social media transparency, asserting that early detection is essential to mitigate the $1.3 billion annual economic loss caused by influencer fraud and advertising waste.
Downloads
References
Akyon, F. C., & Kalfaoglu, E. (2019). Instagram fake and automated account detection. arXiv preprint arXiv:1907.07091. https://arxiv.org/abs/1907.07091
Akyon, F. Ç., & Kalfaoglu, M. (2019). Detecting fake accounts on social media. arXiv preprint arXiv:1910.03090. https://arxiv.org/abs/1910.03090
Anonymous. (2024). Fake account detection using LSTM-based deep learning models across multiple social media platforms. Future Internet, 16(10), 367. https://doi.org/10.3390/fi16100367
Azami, P., & Passi, K. (2024). Detecting fake accounts on Instagram using machine learning and hybrid optimization algorithms. Algorithms, 17(10), 425. https://doi.org/10.3390/a17100425
Azer, M. A., Mohamed, A. A., Hassan, A. A., & El-Moursy, A. A. (2024). Fake accounts detection in online social networks using hybrid machine learning models. ResearchGate. https://www.researchgate.net/publication/370112064_Fake_Accounts_Detection_in_Online_Social_Networks_using_Hybrid_Machine_Learning_Models
BeleRicks11. (2023). Instagram fake account detection [Source code]. GitHub. https://github.com/BeleRicks11/Instagram_Fake_Account_Detection
Bharne, A. P., & Bhaladhare, D. (2023). A machine learning approach for fake account detection on Instagram using textual and visual features. Electronics, 13(8), 1571. https://doi.org/10.3390/electronics13081571
Bordbar, M., Matwin, S., & Abdar, M. (2022). Semi-supervised generative adversarial networks for fraud detection with imbalanced data. arXiv preprint arXiv:2212.01071. https://arxiv.org/abs/2212.01071
Breuer, R., Kloft, M., & Backes, M. (2020). Early detection of fake accounts on social networks using SybilEdge. arXiv preprint arXiv:2004.04834. https://arxiv.org/abs/2004.04834
Celik, E., & Omurca, S. I. (2025). A Novel Framework Leveraging Social Media Insights to Address the Cold-Start Problem in Recommendation Systems. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), 234. https://doi.org/10.3390/jtaer20030234
Fani, H., Rezapour, R., Pournamdarian, A., & Panah, A. A. (2023). Countering social media cybercrime using deep learning: Instagram fake accounts detection. Future Internet, 15(10), 367. https://doi.org/10.3390/fi15100367
Harris, J., Jones, L., & Kim, S. (2023). Detecting fake accounts on Instagram using ensemble machine learning classifiers. Electronics, 13(8), 1571. https://doi.org/10.3390/electronics13081571
JESTER. (2018). An Inferable Representation Learning for Fraud Review Detection Catering for Cold-Start. (2018)
Liu, Y., Wu, Y., & Li, J. (2024). A comprehensive review of deep learning approaches for fake account detection in online social networks. arXiv preprint arXiv:2410.20293. https://doi.org/10.48550/arXiv.2410.20293
omchaudhari01. (2023). Instagram fake account detection using machine learning [Source code]. GitHub. https://github.com/omchaudhari01/Instagram-Fake-Account-Detection-Using-Machine-Learning-
Sallah, M., Hassanein, A., & El-Sayed, A. (2022). Machine learning interpretability to detect fake accounts in Instagram. International Journal of Cloud Applications and Computing, 12(3), 1–15. https://doi.org/10.4018/IJCAC.303665
Verma, P., Yadav, R., & Sharma, A. (2025). Detection of fake Instagram accounts using gradient boosting machine learning model. International Journal of Scientific Research in Science, Engineering and Technology, 12(2), 234–242. https://ijsrset.com/index.php/home/article/view/IJSRST25122234
Winston, L., Omoseebi, A., & Collines, J. (2025). Deep Learning Models for Behavior-Based Fake Account Detection. (June 12, 2025).
Zarei, K., Farahbakhsh, R., & Crespi, N. (2020). A first Instagram dataset on COVID-19. arXiv preprint arXiv:2010.08438. https://doi.org/10.48550/arXiv.2010.08438
Zarei, K., Farahbakhsh, R., Crespi, N., & Tyson, G. (2020). Detecting fake accounts on social networks using deep neural networks: A case study of Instagram. arXiv preprint arXiv:2010.08438. https://arxiv.org/abs/2010.08438
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Lydia Mwanazaire, Dr John Kamau
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution (CC-BY) 4.0 License that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
