Leveraging Deep Learning Approaches for Deepfake Detection: A Review

View on arXiv ← Back to list

Authors: Aniruddha Tiwari, Rushit Dave, Mounika Vanamala

Published: 2023-04-04 16:04:42+00:00

AI Summary

This review paper explores various deep learning methodologies for deepfake detection, aiming to identify cost-effective models with high accuracy and generalizability across different datasets. The authors analyze existing approaches using CNNs, RNNs, and LSTMs, highlighting their strengths and limitations.

Abstract

Conspicuous progression in the field of machine learning and deep learning have led the jump of highly realistic fake media, these media oftentimes referred as deepfakes. Deepfakes are fabricated media which are generated by sophisticated AI that are at times very difficult to set apart from the real media. So far, this media can be uploaded to the various social media platforms, hence advertising it to the world got easy, calling for an efficacious countermeasure. Thus, one of the optimistic counter steps against deepfake would be deepfake detection. To undertake this threat, researchers in the past have created models to detect deepfakes based on ML/DL techniques like Convolutional Neural Networks. This paper aims to explore different methodologies with an intention to achieve a cost-effective model with a higher accuracy with different types of the datasets, which is to address the generalizability of the dataset.

Key findings

The review shows that CNN-based models, especially those incorporating temporal features, generally perform well but often lack generalizability across different datasets. Models focusing on analyzing human behavioral patterns, such as eye-blinking, also demonstrate promising results. However, the creation of increasingly realistic deepfakes remains a challenge for current detection techniques.

Approach

The paper reviews existing deepfake detection methods, analyzing models based on Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), and their combinations. The analysis focuses on the datasets used, model architectures, and achieved accuracy, identifying limitations and suggesting future research directions.

Datasets

Celeb-Deepfake (Celeb-DF), FaceForensics++, FFIW10K, Deepfake Detection Challenge (DFDC), WildDeepfake, UADFV, and others mentioned in the literature review.

Model(s)

DenseNet169, MesoNet, DFDNet, ResNet-50, InceptionV3, XceptionNet, LSTM, InceptionResNetV2, CNN-RNN, Long-Term Recurrent Network (LRCN), and others mentioned in the literature review.

Author countries

USA

← Previous