Improved Accuracy in Deepfake Detection Using GAN and Fisherface Algorithm

M. Udhaya Kumar, B. Latha, B. Vinoth Kumar, R. Srinithi, B. Elamathi, C. Soundharya

2025

Abstract

Aim: The goal of this paper is to design a robust deepfake detection method by integrating GAN, and the Fisherface algorithm to increase accuracy and precision in the detection of fabricated media. The performance of this model is compared with traditional CNN and LSTM-based models. This research is into two groups. Group 1 is a CNN-LSTM with 950 samples in order to capture spatial and temporal features. Group 2 utilizes GANs in synthetic data augmentation with Fisherface feature extraction and SVM classification with 1030 samples. Results: The hybrid GAN-Fisherface-SVM model results in significantly higher detection accuracy compared to traditional models. The hybrid model shows a significant gain of detection, which stands out at about 5-10%, by measurement matrices such as accuracy, error rate and response time with a significance value below 0.05. Conclusion: Overall, the developed approach that combines data augmentation using a GAN method along with a Fisherface algorithm performs a dramatic level of recognition towards deep fake as compared with methods used.

Paper Citation

in Harvard Style

Kumar M., Latha B., Kumar B., Srinithi R., Elamathi B. and Soundharya C. (2025). Improved Accuracy in Deepfake Detection Using GAN and Fisherface Algorithm. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 70-77. DOI: 10.5220/0013877000004919

in Bibtex Style

@conference{icrdicct`2525,
author={M. Kumar and B. Latha and B. Kumar and R. Srinithi and B. Elamathi and C. Soundharya},
title={Improved Accuracy in Deepfake Detection Using GAN and Fisherface Algorithm},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25},
year={2025},
pages={70-77},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013877000004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - ICRDICCT`25
TI - Improved Accuracy in Deepfake Detection Using GAN and Fisherface Algorithm
SN - 978-989-758-777-1
AU - Kumar M.
AU - Latha B.
AU - Kumar B.
AU - Srinithi R.
AU - Elamathi B.
AU - Soundharya C.
PY - 2025
SP - 70
EP - 77
DO - 10.5220/0013877000004919
PB - SciTePress