Identifying Deepfake Videos and Audio Using Artificial Intelligence: A Review

This review paper is authored by Swain Sujeet Dsilva as part of a Research Methodology project it surveys artificial intelligence methods for detecting manipulated video and audio (deepfakes). The paper organizes existing detection approaches into three main categories: Handcrafted feature-based methods (such as blink cues, noise analysis, and cepstral features) Deep learning-based methods (including spatial CNNs, frequency-aware networks, and self-supervised speech encoders) Multimodal fusion methods (combining audio and visual streams at the feature level, decision level, or via cross-modal consistency) In addition, it also summarizes major public benchmark datasets across video, audio, and audio-visual multimedia formats (such as FaceForensics++, Celeb-DF, ASVspoof, and FakeAVCeleb), compares representative evaluation metrics and reported performance, and examines key open challenges including cross-dataset generalization, compression robustness, adversarial vulnerability, and demographic bias alongside future research directions.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-13
DOI
https://doi.org/10.5281/zenodo.24187184
Primary Topic
Digital Media Forensic Detection
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Identifying Deepfake Videos and Audio Using Artificial Intelligence: A Review

Swain Sujeet Dsilva
Zenodo (CERN European Organization for Nuclear Research)
Digital Media Forensic Detection
preprint

Identifying Deepfake Videos and Audio Using Artificial Intelligence: A Review

Swain Sujeet Dsilva
preprint en

Abstract

This review paper is authored by Swain Sujeet Dsilva as part of a Research Methodology project it surveys artificial intelligence methods for detecting manipulated video and audio (deepfakes). The paper organizes existing detection approaches into three main categories: Handcrafted feature-based methods (such as blink cues, noise analysis, and cepstral features) Deep learning-based methods (including spatial CNNs, frequency-aware networks, and self-supervised speech encoders) Multimodal fusion methods (combining audio and visual streams at the feature level, decision level, or via cross-modal consistency) In addition, it also summarizes major public benchmark datasets across video, audio, and audio-visual multimedia formats (such as FaceForensics++, Celeb-DF, ASVspoof, and FakeAVCeleb), compares representative evaluation metrics and reported performance, and examines key open challenges including cross-dataset generalization, compression robustness, adversarial vulnerability, and demographic bias alongside future research directions.

Zenodo (CERN European Organization for Nuclear Research)
Digital Media Forensic Detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon