Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 84 of the 84 references it cites were found in a registry or an index; 17 of the 42 sources its claims cite were checked for support.

A Comprehensive Review of DeepFake Detection Using Advanced Machine Learning and Fusion Methods

Open access
Gupta, Gourav, Raja, Kiran, Gupta, Manish, Jan, Tony, Whiteside, Scott Thompson, Prasad, Mukesh · 2023 · Electronics

Abstract

Recent advances in Generative Artificial Intelligence (AI) have increased the possibility of generating hyper-realistic DeepFake videos or images to cause serious harm to vulnerable children, individuals, and society at large with misinformation. To overcome this serious problem, many researchers have attempted to detect DeepFakes using advanced machine learning techniques and advanced fusion techniques. This paper presents a detailed review of past and present DeepFake detection methods with a particular focus on media-modality fusion and machine learning. This paper also provides detailed information on available benchmark datasets in DeepFake detection research. This review paper addressed the 67 primary papers that were published between 2015 and 2023 in DeepFake detection, including 55 research papers in image and video DeepFake detection methodologies and 15 research papers on identifying and verifying speaker authentication.

Overview

CoChat Summary

TL;DR

This is a survey paper, not an original experiment: it reviews DeepFake detection methods with a focus on modality fusion and advanced machine learning, covering 67 primary papers published from 2015 to 2023. It also summarizes benchmark datasets, reports representative results from prior work, and argues that better cross-dataset generalization, efficiency, and robustness remain open problems.

Key findings

  • The review covers 67 primary papers: 55 on image/video DeepFake detection and 15 on speaker authentication, spanning 2015–2023.
  • It states that benchmark datasets expanded sharply: total DeepFake videos grew from 3,038 in 2018 to 188,154 in 2020, and the DFDC-Full Dataset is reported as the largest corpus, while UADFV is the smallest.
  • The paper reports several prior-method results, including:
  • MesoInception-4 achieved 98% on the DeepFake dataset and 95% on Face2Face.
  • A triplet-loss/Xception-based approach reached AUC 99.2% on Celeb-DF and accuracy 99.71% on a highly compressed neural texture.
  • An eye-blink-based method reported 93.23% and 98.1% accuracy for real and fake videos, respectively, and another approach reported 0.321 log-likelihood error on DFDC.
  • For audio spoof detection, the review cites examples such as TECC with 0.161% average EER on S1–S9 spoofing algorithms and 12 EER on ASVSpoof2017 for a feature ensemble, and notes that some methods improve t-DCF and EER on ASVSpoof2019.

Methods

  • This is a narrative literature review of DeepFake detection, organized around benchmark datasets, video/image modality fusion, audio modality fusion, and advanced DeepFake detection methods.
  • The sample is the set of 67 primary papers cited and discussed in the review; the paper also reproduces dataset summary tables for DeepFake and audio spoofing benchmarks.
  • The main analysis is comparative synthesis of prior work, including reported metrics and method families such as CNNs, RNNs/LSTMs, attention, capsule networks, transfer learning, multimodal fusion, and benchmark evaluation.

Limitations

  • The paper states that DeepFake detection models still have difficulty generalizing across diverse datasets, with reduced effectiveness under changes in lighting, facial expressions, and video quality.
  • It also states that an unseen class in the testing dataset relative to the training dataset is a major challenge, and that performance degrades when the testing set differs significantly from the training set.
  • Evident from the text: this review does not present a new dataset, new benchmark, or new experimental model of its own; it summarizes prior studies and reports their published results.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/electronics13010095

Cites

84 of 84 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

15 Valid · 10 Not valid · 23 Not checked · 20 Own results and interpretations

9 more claims

Create a free account to see all 68 claims and to verify the 5 references still open.

See all claims — free account
  • Probably not a substantial claim. 37
  • It may state two separate claims as one. 29

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

1 of 1

Citations

Cites

Show all 76 references

Research works printed but not linked: 2

  1. Symeon, P.C.G.K.Z., and Kompatsiaris, P.I. (2020). AFace PREPROCESSING APPROACH FOR IMPROVED DEEPFAKE DETECTION. arXiv.
  2. Sohrawardi, S.J., Seng, S., Chintha, A., Thai, B., Hickerson, A., Ptucha, R., and Wright, M. (2020, January 20–21). Defaking DeepFakes: Understanding journalists’ needs for DeepFake detection. Proceedings of the Computation+ Journalism 2020 Conference, Boston, MA, USA.

Cited by

See all 99

Metadata

Authors
Gupta, G., Raja, K., Gupta, M., Jan, T., Whiteside, S. T., & Prasad, M.
Affiliations
  • Norwegian University of Science and Technology (NO)
  • Torrens University Australia (AU)
  • Artificial Intelligence and Optimization Research Centre, Design and Creative Technology, Torrens University, 46 Mountain Street, Ultimo, NSW 2007, Australia
  • University of Technology Sydney (AU)
Corresponding author
Gupta, Manish
Venue
Electronics
Publisher
MDPI AG
Published
2023-12-25
Volume
13
Issue
1
Pages
95
ISSN
2079-9292 (ISSN-L)
OpenAlex
W4390235290
Cited by
99
References
76
Citations by year
2026: 33 · 2025: 47 · 2024: 19
FWCI
9.00
Type
journal article
Language
en
Open access
Locations
Electronics (publishedVersion, cc-by)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
Keywords
deepfake detection, generative adversarial networks, multimodal fusion, misinformation, speaker verification, generative AI

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.