Downloads: 1
Research Paper | Computer Science and Engineering | Volume 15 Issue 7, July 2026 | Pages: 1781 - 1790 | India
Hierarchical Multimodal Disinformation Detection System for Armed Conflict: Integration of Forensic, Geopolitical, and Social Intelligence
Abstract: The proliferation of weaponized disinformation during armed conflicts poses an unprecedented threat to global stability. This paper presents a comprehensive 10-module hierarchical multimodal framework designed specifically to detect disinformation in high-stakes military scenarios. The architecture integrates an 8-module video analysis pipeline, an optimal 3-level hierarchical image ensemble, and a dynamic social media narrative propagation analyzer. The video framework utilizes a two-layer fusion strategy- combining parameter-free geometric-mean forensic fusion with a data-validated decision tree?achieving 100% classification accuracy and a 0% false-positive rate on a curated pilot benchmark of 30 videos. The Image Authenticity module integrates CNN forensics, CLIP-based image-caption consistency, and LLM visual reasoning to overcome adversarial vulnerabilities, achieving 96-100% operational accuracy on a 25-image pilot set. Concurrently, the Social Media Impact module processes complex multilingual narrative networks using a novel True Impact Score (TIS), yielding a 25% AUROC improvement in detecting coordinated campaigns. While general-purpose models exhibit severe domain degradation (22-44%) on military content, this domain-specific architecture eliminates degradation through the integration of a novel 93-feature Geopolitical Trust Score and semantic verification. Given the limited size of the pilot benchmarks, these results should be read as proof-of-concept evidence rather than population-level performance estimates; a mathematically grounded validation protocol (N = 385) is outlined to formally quantify real-world variance and operational margins ahead of edge deployment.
Keywords: Armed Conflicts, Geopolitical Intelligence, Hierarchical Fusion, Image Authenticity, LLM Verification, Mul-timodal Disinformation Detection, Narrative Propagation, True Impact Score
How to Cite?: Lakhan Singh, "Hierarchical Multimodal Disinformation Detection System for Armed Conflict: Integration of Forensic, Geopolitical, and Social Intelligence", Volume 15 Issue 7, July 2026, International Journal of Science and Research (IJSR), Pages: 1781-1790, https://www.ijsr.net/getabstract.php?paperid=SR26721152047, DOI: https://dx.doi.org/10.21275/SR26721152047