Guarding Authenticity: A Deepfake Detection System

Guarding Authenticity: A Deepfake Detection System

Prof. N. R. Thorat1, Prajwal Mohite2, Anubhaw Mishra3, Rohit Sawant4

1Assistant Professor, Department of Computer Engineering, SITS, Pune

2BE Graduate (IV year), Department of Computer Engineering, SITS, Pune

3BE Graduate (IV year), Department of Computer Engineering, SITS, Pune

4BE Graduate (IV year), Department of Computer Engineering, SITS, Pune

---------------------------------------------------------------------***---------------------------------------------------------------------

ABSTRACT -

This Recent advancements in computer vision have led to the development of powerful tools that can create realistic deepfakes. A generative adversarial network (GAN) can manipulate captured media streams, such as images, audio, and video, to make them appear to fit other environments. The spread of these fake media streams can cause chaos in social communities and damage the reputation of individuals or groups. It can also influence public sentiments and opinions toward the targeted person or community. Researchers have suggested using convolutional neural networks (CNNs) as an effective method for detecting deepfakes in the network. However, most existing techniques struggle to capture the dissimilarities between frames in the collected media streams. Motivated by this challenge, this paper presents a novel and improved deep-CNN (D-CNN) architecture for deepfake detection. The proposed approach aims to achieve reasonable accuracy and high generalizability. The model is trained on images from multiple sources, which enhances its overall generalizability capabilities.A binary-cross entropy and Adam optimizer are utilized to improve the learning rate of the D-CNN mode.

Key Words:      cnn, deepfake, image, network, media stream, detection