International Journal of Scientific Research and Engineering Development

Deepfakes, or supremely realistic synthetic media, are an outcome of the rapid growth of deep learning. Despite the impressive progress that these technologies are making in picture synthesis and face alteration, they also bring up serious concerns regarding cybersecurity, identity theft, intrusion of privacy and fake news. To overcome these challenges, this paper proposes Deepfake Detection Platform for Social Media, a web application powered by AI, which can immediately recognize whether uploaded photos are authentic or not. The proposed platform is a full-stack architecture that uses React.js on the frontend and Node.js/Express.js on the backend and MongoDB on the database with a system- optimized ResNet-18 convolutional neural network. To enhance robustness, the detection model is trained on a mixture of both real and re-edited photos collected on publicly available deepfake datasets, including FaceForensics++ and DFDC, and synthetic ones. A preprocessing pipeline consists of frame extraction, face detection, resizing, normalization, and augmentation to enhance the quality of training and generalization of the model. Experimental results show that the model has a low inference latency, which makes it applicable in realtime, and can correctly classify real and fake images. There is also the Docker-based deployment that will allow scaling, history of uploads, and secure authentication. The proposed platform demonstrates how deep learning can be transformed into a practical, easy-to-use means of combating fake information on social media.

Shifa Bilal Tamboli, Simeen Phiroj Mulani, Arman Tajuddin Shiakh,"Deepfake Detection in Social Media Using Fine-Tuned ResNet-18 and Full-Stack Deployment" International Journal of Scientific Research and Engineering Development, V9(2): Page(2335-2345) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.