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{195132,
        author = {M.R.K Raju and Ch.Harika and Ch.S.D.Prem Kumar and B.M.L.Sudha and G.Iswarya},
        title = {Deepfake Detection: Deep Learning Based System for Identifying Synthetic Media},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {10},
        pages = {8194-8199},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=195132},
        abstract = {The rapid advancement of Artificial Intelligence (AI) and Deep Learning (DL) has significantly transformed digital media generation. Among these developments, deepfake technology has gained considerable attention due to its ability to create highly realistic manipulated images and videos. Deepfakes are generated using deep neural networks that can replace faces, modify expressions, or alter speech patterns with convincing accuracy. Although this technology has useful applications in entertainment, filmmaking, and virtual environments, its misuse for misinformation, identity fraud, cyber harassment, and political manipulation has raised serious ethical and security concerns. This has created an urgent need for reliable deepfake detection systems. 
This research proposes a deep learning-based framework for identifying manipulated media using a hybrid CNN-LSTM architecture. The model employs a pretrained MobileNetV2 Convolutional Neural Network (CNN) to extract spatial features from image frames, while a Long Short-Term Memory (LSTM) network captures temporal dependencies across video sequences. The system processes both images and videos by extracting twenty representative frames, resizing them to 224*224 pixel, and applying normalization before classification. The dataset used for training and evaluation includes balanced samples from the Celeb-DF, and Deepfake Detection Challenge (DFDC) datasets. The model was implemented using TensorFlow and trained with GPU support in Google Collab. Experimental results indicate an approximate accuracy of 80%, demonstrating effective discrimination between real and manipulated content. A Gradio-based interface was also developed to enable practical media verification.},
        keywords = {Deepfake Detection, CNN-LSTM, Synthetic Media, Temporal Analysis, Digital Forensics},
        month = {March},
        }