Deep fake Video Detection Using Deep Learning

Abstract

In recent years, the availability of free deep learning tools has significantly simplified the creation of highly realistic face-swapped videos, commonly known as DeepFake (DF) videos. Although video manipulation techniques have existed for many years, recent advancements in deep learning have greatly enhanced the realism of such content while also making it more accessible to generate. These AI-synthesized videos can now be produced with ease using various artificial intelligence-based applications. Despite their growing prevalence, detecting DeepFake videos remains a complex challenge. Developing reliable algorithms for their identification is difficult due to the high level of visual realism. To address this issue, a detection system based on Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks has been proposed. In this approach, the CNN is used to extract spatial features from individual video frames, which are then fed into an LSTM-based Recurrent Neural Network (RNN) to capture temporal dependencies across frames and determine whether the video is authentic or manipulated. The model focuses on identifying temporal inconsistencies that are often introduced during the DeepFake generation process. The proposed system has been evaluated on a large dataset of manipulated videos and demonstrates competitive performance, even with a relatively simple architecture, highlighting its effectiveness in detecting DeepFake content.