Recent Advancements In The Field Of Deepfake Detection

View on arXiv ← Back to list

Authors: Natalie Krueger, Mounika Vanamala, Rushit Dave

Published: 2023-08-10 13:24:27+00:00

AI Summary

This survey paper analyzes recent advancements in deepfake detection, focusing on neural network-based methods and their effectiveness across various modalities (audio, video, and joint audio-video). It also examines the impact of datasets and the limitations of existing approaches.

Abstract

A deepfake is a photo or video of a person whose image has been digitally altered or partially replaced with an image of someone else. Deepfakes have the potential to cause a variety of problems and are often used maliciously. A common usage is altering videos of prominent political figures and celebrities. These deepfakes can portray them making offensive, problematic, and/or untrue statements. Current deepfakes can be very realistic, and when used in this way, can spread panic and even influence elections and political opinions. There are many deepfake detection strategies currently in use but finding the most comprehensive and universal method is critical. So, in this survey we will address the problems of malicious deepfake creation and the lack of universal deepfake detection methods. Our objective is to survey and analyze a variety of current methods and advances in the field of deepfake detection.

Key findings

Neural network-based methods, particularly those focusing on identity, features, and emotion detection, show high accuracy in deepfake detection. However, dataset biases and limitations in handling various video qualities and contexts remain significant challenges. Human performance in deepfake detection varies considerably and is often comparable to or surpassed by machine learning models in certain contexts.

Approach

The paper surveys existing deepfake detection methods, primarily focusing on neural network architectures (CNNs, RNNs, LSTMs, Vision Transformers) applied to various features like audio-visual synchronization, facial features, and emotional inconsistencies. It also explores non-neural network methods and the effects of dataset biases.

Datasets

FaceForensics++, DFDC, Celeb-DF, TIMIT-DF, Vox-DeepFake, GBDF, DeePhy

Model(s)

CNNs, RNNs, LSTMs, Vision Transformers, YOLO, SpecRNet, EfficientNet B7, Xception, EfficientNet B4

Author countries

USA

← Previous