DeepFace Detection Using Recurrent Neural Networks

In recent months, the proliferation of machine learning-based software tools has facilitated the creation of convincing face swaps in videos, resulting in what are commonly referred to as “deepfake” videos. These videos can be highly deceptive, leaving minimal traces of manipulation and posing significant risks in various scenarios, including political manipulation, blackmail, and the fabrication of terrorism events.To address this challenge, this paper proposes a temporal-aware pipeline for the automatic detection of deepfake videos. Our approach utilizes a convolutional neural network (CNN) to extract frame-level features from the videos. These features are then employed to train a recurrent neural network (RNN), enabling the model to classify whether a video has undergone manipulation or not.We evaluate our method using a diverse dataset of deepfake videos collected from multiple online sources. Our results demonstrate the effectiveness of our approach, achieving competitive performance in detecting manipulated videos while employing a simple architecture.By leveraging the temporal information present in videos, our system offers a robust solution for identifying deepfake content, thereby mitigating the potential risks associated with the proliferation of deceptive media..