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{201255,
        author = {Yuvraj Singh Pawar and Rushikesh Agarwal and Soham Dahatonde and Dr. Anant Kaulage},
        title = {Deepfake Detection Using ML},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {3177-3183},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=201255},
        abstract = {The rapid development in generative artificial intelligence has dramatically improved the realism and availability of synthetic media, raising serious concerns about misinformation, identity theft, and the authenticity of digital content. Traditional deep learning-based detection methods mainly rely on spatial features extracted from images, and often neglect artifacts in the frequency domain [3], [4].
In this work, we propose a hybrid deepfake detection framework by combining the convolutional neural network (CNN)-based spatial feature extraction with the frequency domain using the Fast Fourier Transform (FFT). The proposed system is based on an EfficientNet-B0 architecture pre-trained on ImageNet and fine-tuned for the binary classification of real and AI-generated images [5]. In parallel, frequency-domain features are analyzed in order to catch high-frequency inconsistencies often seen in synthetic media [6].
The system incorporates Grad-CAM based visualization to provide better interpretability by identifying the regions of an image that influence the model’s prediction [7]. We also evaluate robustness under real-world distortions such as compression, noise, and blur.
The system is deployed on a FastAPI-based backend, enabling inference on images and video frames in real-time. Experimental results show that the hybrid approach can improve the detection reliability and generalization ability in comparison with the individual CNN models, and provide explainable information for the model decision.},
        keywords = {Deepfake Detection, EfficientNet, Frequency Domain Analysis, FFT, Explainable AI, Computer Vision, Image Forensics},
        month = {May},
        }