Citation

Mohammed, Bahaa Kareem and Alogaili, Riyadh Rahef Nuiaa and Hasan, Baraa Mohammed and Dashoor, Zeinab Ali and Alshami, Haider Ghazi and Manickam, Selvakumar and Arjuman, Navaneethan C. (2025) AMGET: An Adaptive Multimodal Graph-Enhanced Transformer for Cybersecure Detection of Evolving Fake News on Social Media. International Journal of Intelligent Engineering and Systems, 18 (10). pp. 166-183. ISSN 2185-310X

Text 14710.pdf - Published Version Restricted to Repository staff only Download (644kB)

Abstract

The proliferation of fake news on social media poses a significant cybersecurity threat, undermining trust and exacerbating public risk. Existing detection methods often overlook critical multimodal content, temporal dynamics, and adaptive learning requirements. To address these gaps, this study introduces the Adaptive Multimodal Graph-Enhanced Transformer (AMGET), a novel architecture synergistically integrating multimodal fusion, temporal heterogeneous graph modeling, and continual learning mechanisms. We develop an innovative cross-modal attention approach, enhancing semantic modality alignment. A temporal graph neural network with decaying attention coefficients captures evolving misinformation spread, while our continual learning mechanism ensures ongoing adaptability against emerging threats. Comprehensive evaluation on the extensive TruthSeeker dataset demonstrates that AMGET substantially surpasses baseline approaches, achieving 96.7% accuracy, 0.961 macro F1-score, and exceptional temporal robustness (87.4% TRI). Crucially, AMGET also provides interpretable insights through multimodal explanations aligning closely with expert reasoning. Our results underline the practical efficacy of AMGET, setting new benchmarks in cybersecurity-oriented fake news detection and significantly advancing multimodal misinformation modeling in dynamic social media contexts.

Item Type:	Article
Uncontrolled Keywords:	Cybersecurity-driven social media analysis
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Nor Afiqah Mohd Adnan
Date Deposited:	06 Nov 2025 03:35
Last Modified:	06 Nov 2025 14:31
URII:	http://shdl.mmu.edu.my/id/eprint/14710

Downloads

Edit (login required)