Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{209909,
        author = {Ashish K. Shrivastva and Prof. Anmol S. Budhewar and Amruta S. Patil and Kartik B. Kawale and Bhumika P. Jadhav},
        title = {A Lightweight Quality-Aware and Explainable Deepfake Detection Framework for Video Verification},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {5},
        pages = {3883-3892},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=209909},
        abstract = {Deepfake videos threaten the reliability of visual evidence since modern face manipulation can co-exist with ordinary degradations such as blur, poor illumination, compression, occlusion, small facial regions and repeated frames. In this paper, we propose a framework-oriented design for lightweight and explainable video verification instead of reporting the unmeasured detection results. The framework integrates IFQA, adaptive key-frame selection, landmark-based face alignment, attention-enhanced EfficientNetV2-B0 classifier, quality-weighted video-level confidence aggregation, and Grad-CAM visualisation. IFQA explicitly scores candidate frames on sharpness, illumination, contrast, face visibility, face-area ratio, and compression quality, and discards low-quality and near-duplicate frames before expensive classification. The remaining facial crops are classified independently and fused with normalised quality weights, without the complexity of a recurrent temporal model. The paper formalises the working stages, equations, implementation workflow, leakage-resistant dataset protocol, planned ablation and robustness evaluation, computational analysis and web-based verification layer. The resulting design targets a clear trade-off between detection reliability, computational efficiency, interpretability and deployment simplicity.},
        keywords = {deepfake detection, video verification, frame quality assessment, EfficientNetV2, explainable artificial intelligence},
        month = {October},
        }