A Machine Learning Approach for DeepFake Detection

View on arXiv ← Back to list

Authors: Gustavo Cunha Lacerda, Raimundo Claudio da Silva Vasconcelos

Published: 2022-09-28 02:46:04+00:00

AI Summary

This paper proposes a DeepFake detection method using convolutional neural networks (CNNs) and the Celeb-DF dataset. The model achieves a 95% accuracy in classifying DeepFake images, demonstrating its effectiveness in detecting manipulated facial images.

Abstract

With the spread of DeepFake techniques, this technology has become quite accessible and good enough that there is concern about its malicious use. Faced with this problem, detecting forged faces is of utmost importance to ensure security and avoid socio-political problems, both on a global and private scale. This paper presents a solution for the detection of DeepFakes using convolution neural networks and a dataset developed for this purpose - Celeb-DF. The results show that, with an overall accuracy of 95% in the classification of these images, the proposed model is close to what exists in the state of the art with the possibility of adjustment for better results in the manipulation techniques that arise in the future.

Key findings

The model achieved an overall accuracy of 95% in classifying DeepFake images. The results, while slightly below state-of-the-art, are promising and show potential for improvement through further fine-tuning and data augmentation. The model demonstrates a high precision in identifying real images.

Approach

The authors used a pre-trained EfficientNet-B4 CNN model, fine-tuned it on the Celeb-DF v2 dataset. The process involved pre-processing videos to extract and normalize facial images, then training and validating the model for binary classification (real/fake).

Datasets

Celeb-DF v2

Model(s)

EfficientNet-B4

Author countries

Brazil

← Previous