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{175349,
        author = {Pavan Bhandekar and Prof. A.B.Deshmukh and Sanika Tole and Achal Surandase and Swaraj P. Patil and Harshit M. Pande},
        title = {DeepFake Detection Using MobileNetV2 and LSTM},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {11},
        pages = {3080-3086},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=175349},
        abstract = {The spread of DeepFake technology threatens digital media integrity to a large extent, calling for effective detection methods. This paper introduces a DeepFake detection system that uses MobileNetV2 for spatial feature extraction, LSTM for temporal analysis, and MTCNN for face detection, with a test accuracy of 95%. We trained the model on the Celeb-DF dataset, which comprises 199 videos (99 real, 100 fake), with 5 frames per video to strike a balance between computational efficiency and detection accuracy. We improved the performance of the model by iterative threshold optimization, increasing accuracy from 85.71% (threshold 0.5) to 95% (threshold 0.18). Our approach involves fine-tuning MobileNetV2 with the addition of temporal analysis using LSTM and optimal thresholding of classification to trade-off between false positives and false negatives. Our experiments showcase the efficacy of our method at detecting nuanced DeepFakes while preserving a high accuracy on natural videos, thus rendering it an effective solution for practical applications.},
        keywords = {Celeb-DF Dataset, DeepFake Detection, LSTM, MobileNetV2, MTCNN, Threshold Optimization.},
        month = {April},
        }