EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Attention-Driven CNN-LSTM Fusion for Robust Deepfake Detection in Digital Media

S. E. Hammed, S. Al-Darraji

AFRICAN JOURNAL OF APPLIED RESEARCH · 15.04.2026 · 12 (3) · 18-56

DOI: 10.26437/z15r9q80 ↗

Kurz gefasst

Fachartikel in der Zeitschrift AFRICAN JOURNAL OF APPLIED RESEARCH, online erschienen am 15.04.2026. Verfasst von S. E. Hammed und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

This paper aims to address the growing challenge of deepfake detection, driven by the increasing impact of synthetic media on digital integrity, privacy, and security. Design/Methodology/ Approach: The proposed approach integrates a hybrid deep learning architecture combining Convolutional Neural Networks (CNNs) to extract spatial features and Long Short-Term Memory (LSTM) to model temporal relationships, enhanced by an attention mechanism to focus on important features and subtle manipulation patterns. The methodology includes video preprocessing such as frame extraction, face detection, alignment, and normalisation, followed by sequence-level classification. Research Limitation: The study is limited by its reliance on benchmark datasets, which may not fully represent real-world scenarios, and by potential challenges in generalising to unseen manipulations. Additionally, no funding support was reported.

Findings

The model is evaluated on the FaceForensics++ dataset using standard metrics, including accuracy, precision, recall, F1-score, and ROC-AUC, demonstrating improved performance in detecting deepfake videos. Practical Implication: The proposed model can be applied in security systems, social media platforms, and digital forensics to detect and prevent the spread of manipulated video content. Social Implication: The work contributes to reducing misinformation, enhancing trust in digital media, and protecting user privacy and societal security. Originality/Value: The originality of this work lies in integrating CNN, LSTM, and attention mechanisms into a unified framework for spatiotemporal feature learning, providing a scalable and effective solution for deepfake detection.

Zitieren

APA

Hammed, S. E., & Al-Darraji, S. (2026). Attention-Driven CNN-LSTM Fusion for Robust Deepfake Detection in Digital Media. AFRICAN JOURNAL OF APPLIED RESEARCH, 12(3), 18-56. https://doi.org/10.26437/z15r9q80

Vancouver

Hammed SE, Al-Darraji S. Attention-Driven CNN-LSTM Fusion for Robust Deepfake Detection in Digital Media. AFRICAN JOURNAL OF APPLIED RESEARCH. 2026;12(3):18-56. doi:10.26437/z15r9q80

BibTeX
@article{Hammed2026,
  author = {Hammed, S. E. and Al-Darraji, S.},
  title = {Attention-Driven CNN-LSTM Fusion for Robust Deepfake Detection in Digital Media},
  journal = {AFRICAN JOURNAL OF APPLIED RESEARCH},
  year = {2026},
  volume = {12},
  number = {3},
  pages = {18--56},
  doi = {10.26437/z15r9q80}
}
Kostenlos starten →