Deepfake Image Detection on Social Media: A Review

In social media, deepfake pictures are rapidly circulated to threaten privacy, security, and reliance on the images. Detectors display very high accuracy on individual datasets but do not demonstrate how they function on unseen generators or recompressed images, nor do they explain their decisions. We review 19 recent papers on image, video,audio, and medical-image detection and group them by method: end-to-end CNN classifiers, hybrid handcrafted–deep representations, consistency and context cues, explainable and localizing detectors, and temporal or multimodal detectors. Our review reveals eight specific issues that require further research support from evidence, among which are unseen-generator generalisation, reality-grounded media transformations, faithfulness to explanation, and so forth. To resolve five of these issues, we put forward a pure-image pipeline that assesses hand-crafted (LBP, LTP, HOG), fine-tuned CNN, and hybrid on identical leakage-safe splits, adds validation-based calibration, and treats explanations with a masking test. The proposal has no data tested; it is merely the design.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24321181
Primary Topic
Digital Media Forensic Detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Deepfake Image Detection on Social Media: A Review

S. B. Bagal, SACHIN GULABRAO BAGUL, Vishal Chaudhari, Rajendra Potdar
3 links
Zenodo (CERN European Organization for Nuclear Research)
Digital Media Forensic Detection
article

Deepfake Image Detection on Social Media: A Review

S. B. Bagal, SACHIN GULABRAO BAGUL, Vishal Chaudhari, Rajendra Potdar
article en

Abstract

In social media, deepfake pictures are rapidly circulated to threaten privacy, security, and reliance on the images. Detectors display very high accuracy on individual datasets but do not demonstrate how they function on unseen generators or recompressed images, nor do they explain their decisions. We review 19 recent papers on image, video,audio, and medical-image detection and group them by method: end-to-end CNN classifiers, hybrid handcrafted–deep representations, consistency and context cues, explainable and localizing detectors, and temporal or multimodal detectors. Our review reveals eight specific issues that require further research support from evidence, among which are unseen-generator generalisation, reality-grounded media transformations, faithfulness to explanation, and so forth. To resolve five of these issues, we put forward a pure-image pipeline that assesses hand-crafted (LBP, LTP, HOG), fine-tuned CNN, and hybrid on identical leakage-safe splits, adds validation-based calibration, and treats explanations with a masking test. The proposal has no data tested; it is merely the design.

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