Enhancing Social Media Security: LSTM-Based Deep Fake Video Detection

The rapid advancement of generative artificial intelligence (AI) techniques has led to the widespread creation and dissemination of deepfake videos, which convincingly manipulated media, containing fabricated content often indistinguishable from reality. Detecting such deepfake videos is a critical challenge in ensuring the originality and credibility of information in the digital age. In this paper, an approach for detecting deepfake videos using generative AI-based methods has been proposed. For this, this paper introduces a deepfake detection method using Long-short term memory (LSTM) based model. Celeb-DF (v2) has been used for this experiment. From the videos in the dataset, faces were extracted, cropped and were saved in a new video having only face images. By using ResNext, a CNN-based approach, feature extraction was performed and classification was done using the LSTM model. Highest accuracy of 95.33% was achieved using this model.