Temporal anomaly detection API for enhanced deepfake video identification: A ResNEXT-LSTM approach
Nature Journal of Emerging Sciences Technologies and Innovations · 28.02.2026 · 5 (3) · 233-251
DOI: 10.65752/k8fx5162 ↗Kurz gefasst
Fachartikel in der Zeitschrift Nature Journal of Emerging Sciences Technologies and Innovations, online erschienen am 28.02.2026. Verfasst von Abdullahi Kabir Muhammed und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Deepfake videos, also ultra-realistic AI-generated or manipulated videos pose a significant threat to the integrity of digital information by enabling misinformation, identity impersonation, and financial fraud. This paper proposes a temporal anomaly detection framework for deepfake video identification based on a hybrid ResNeXt convolutional neural network and Long Short-Term Memory (LSTM) architecture. The ResNeXt network extracts discriminative spatial facial features from video frames, while the LSTM models temporal dependencies to identify subtle inconsistencies characteristic of deepfake manipulations. The proposed model is trained and evaluated on the Deepfake Detection Challenge dataset, achieving an overall detection accuracy of approximately 98%, with strong precision and recall. To support practical applicability, the framework is implemented as a RESTful API with an accompanying web-based interface, enabling near real-time deepfake detection from uploaded video content. The system is designed for scalable deployment and seamless integration into media verification platforms, digital forensic workflows, and cybersecurity monitoring systems. Experimental results demonstrate that the joint modeling of spatial and temporal information significantly enhances deepfake detection performance, providing an effective and deployable solution for safeguarding trust in digital video content.
Zitieren
APA
Muhammed, A. K., Adamu-Fika, F., & Ifeanyi, F. (2026). Temporal anomaly detection API for enhanced deepfake video identification: A ResNEXT-LSTM approach. Nature Journal of Emerging Sciences Technologies and Innovations, 5(3), 233-251. https://doi.org/10.65752/k8fx5162
Vancouver
Muhammed AK, Adamu-Fika F, Ifeanyi F. Temporal anomaly detection API for enhanced deepfake video identification: A ResNEXT-LSTM approach. Nature Journal of Emerging Sciences Technologies and Innovations. 2026;5(3):233-251. doi:10.65752/k8fx5162
BibTeX
@article{Muhammed2026,
author = {Muhammed, Abdullahi Kabir and Adamu-Fika, Fatimah and Ifeanyi, Fortune},
title = {Temporal anomaly detection API for enhanced deepfake video identification: A ResNEXT-LSTM approach},
journal = {Nature Journal of Emerging Sciences Technologies and Innovations},
year = {2026},
volume = {5},
number = {3},
pages = {233--251},
doi = {10.65752/k8fx5162}
}
EUVIMED