Generative artificial intelligence and synthetic media: A multimodal framework for detection, trust, and authenticity
International Journal of Applied Resilience and Sustainability · 07.06.2026 · 2 (3) · 306-323
DOI: 10.70593/deepsci.0203012 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Applied Resilience and Sustainability, online erschienen am 07.06.2026. Verfasst von Hari Krishna Silamanthula und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
It is now possible to generate images, audio, video, and text which appear and sound authentic using generative artificial intelligence. This artificial media travels at a high velocity through social media. Normal users cannot be consistent as to whether it is actual content or a fabricated content. The majority of detectors that exist today analyze individual types of data. They would lose their precision in cases where one content combines dozens of modalities. Nor do they resist in little, careful assaults. These relationships of trust between individuals and detection tools remain tenuous and the provenience trail on the origin of content is usually absent. The problem of this problem is made up. The goal is to develop and test conceptually a multimodal structure that looks at detection, trust, and authenticity as a unit. The approach is simulation and modeling. We conduct a Monte Carlo simulation on synthetically simulated detection scores of four modalities: visual, audio, text and provenance metadata. This modality scores are used to produce reliability-weighted late fusion. The fused features are then associated with the probability of the proper classification of authenticity using logistic regression. Trust is calculated by calibration error expected, and an explainability term. Simulated results indicate that reliability-weighted multimodal fusion increases mean accuracy and area under the curve compared to each single-modality SARA. Fusion further reduces the calibration error and is much resistant to adversarial perturbation. The most related variables to increased trust are provenance strength and explainability.
Zitieren
APA
Silamanthula, H. K., Molagavalli, R., Singavarapu, R. B., & Mallavarapu, S. S. R. (2026). Generative artificial intelligence and synthetic media: A multimodal framework for detection, trust, and authenticity. International Journal of Applied Resilience and Sustainability, 2(3), 306-323. https://doi.org/10.70593/deepsci.0203012
Vancouver
Silamanthula HK, Molagavalli R, Singavarapu RB, Mallavarapu SSR. Generative artificial intelligence and synthetic media: A multimodal framework for detection, trust, and authenticity. International Journal of Applied Resilience and Sustainability. 2026;2(3):306-323. doi:10.70593/deepsci.0203012
BibTeX
@article{Silamanthula2026,
author = {Silamanthula, Hari Krishna and Molagavalli, Rajesh and Singavarapu, Raja Bhushanam and Mallavarapu, S Sankara Rao},
title = {Generative artificial intelligence and synthetic media: A multimodal framework for detection, trust, and authenticity},
journal = {International Journal of Applied Resilience and Sustainability},
year = {2026},
volume = {2},
number = {3},
pages = {306--323},
doi = {10.70593/deepsci.0203012}
}
EUVIMED