Enhanced Deepfake Detection Using Advanced CNN’s

Abstract

Deepfake by developing a hybrid paradigm for deep learning that combines ResNet50 with LSTM networks for detecting manipulated media. The model analyses both temporal and spatial characteristics in video content, achieving 94.63% validation accuracy on DFDC, or the Deepfake Detection Challenge dataset. Implemented in Python using Kera’s and other ML libraries, this solution provides an effective, scalable approach to combat harmful deepfake content and misinformation.