Enregistré dans:
Détails bibliographiques
Auteurs principaux: Chandra, Nuria Alina, Lee, Hannah, Murtfeldt, Ryan, Qiu, Lin, Karmakar, Arnab, Tanumihardja, Emmanuel, Farhat, Kevin, Caffee, Ben, Lee, Changyeon, Choi, Jongwook, Paik, Sejin, Kim, Aerin, Etzioni, Oren
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2503.02857
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913164785876992
author Chandra, Nuria Alina
Lee, Hannah
Murtfeldt, Ryan
Qiu, Lin
Karmakar, Arnab
Tanumihardja, Emmanuel
Farhat, Kevin
Caffee, Ben
Lee, Changyeon
Choi, Jongwook
Paik, Sejin
Kim, Aerin
Etzioni, Oren
author_facet Chandra, Nuria Alina
Lee, Hannah
Murtfeldt, Ryan
Qiu, Lin
Karmakar, Arnab
Tanumihardja, Emmanuel
Farhat, Kevin
Caffee, Ben
Lee, Changyeon
Choi, Jongwook
Paik, Sejin
Kim, Aerin
Etzioni, Oren
contents In the age of increasingly realistic generative AI, robust deepfake detection is essential for mitigating fraud and disinformation. While many deepfake detectors report high accuracy on academic datasets, we show that these academic benchmarks are out of date and not representative of real-world deepfakes. We introduce Deepfake-Eval-2024, a new deepfake detection benchmark consisting of in-the-wild deepfakes collected from social media and deepfake detection platform users in 2024. Deepfake-Eval-2024 consists of 45 hours of videos, 56.5 hours of audio, and 1,975 images, encompassing the latest manipulation technologies. The benchmark contains diverse media content from 88 different websites in 52 different languages. We find that the performance of open-source state-of-the-art deepfake detection models drops precipitously when evaluated on Deepfake-Eval-2024, with AUC decreasing by 50% for video, 48% for audio, and 45% for image models compared to previous benchmarks. We also evaluate commercial deepfake detection models and models finetuned on Deepfake-Eval-2024, and find that they have superior performance to off-the-shelf open-source models, but do not yet reach the accuracy of deepfake forensic analysts. The dataset is available at https://github.com/nuriachandra/Deepfake-Eval-2024.
format Preprint
id arxiv_https___arxiv_org_abs_2503_02857
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of Deepfakes Circulated in 2024
Chandra, Nuria Alina
Lee, Hannah
Murtfeldt, Ryan
Qiu, Lin
Karmakar, Arnab
Tanumihardja, Emmanuel
Farhat, Kevin
Caffee, Ben
Lee, Changyeon
Choi, Jongwook
Paik, Sejin
Kim, Aerin
Etzioni, Oren
Computer Vision and Pattern Recognition
Artificial Intelligence
Computers and Society
In the age of increasingly realistic generative AI, robust deepfake detection is essential for mitigating fraud and disinformation. While many deepfake detectors report high accuracy on academic datasets, we show that these academic benchmarks are out of date and not representative of real-world deepfakes. We introduce Deepfake-Eval-2024, a new deepfake detection benchmark consisting of in-the-wild deepfakes collected from social media and deepfake detection platform users in 2024. Deepfake-Eval-2024 consists of 45 hours of videos, 56.5 hours of audio, and 1,975 images, encompassing the latest manipulation technologies. The benchmark contains diverse media content from 88 different websites in 52 different languages. We find that the performance of open-source state-of-the-art deepfake detection models drops precipitously when evaluated on Deepfake-Eval-2024, with AUC decreasing by 50% for video, 48% for audio, and 45% for image models compared to previous benchmarks. We also evaluate commercial deepfake detection models and models finetuned on Deepfake-Eval-2024, and find that they have superior performance to off-the-shelf open-source models, but do not yet reach the accuracy of deepfake forensic analysts. The dataset is available at https://github.com/nuriachandra/Deepfake-Eval-2024.
title Deepfake-Eval-2024: A Multi-Modal In-the-Wild Benchmark of Deepfakes Circulated in 2024
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computers and Society
url https://arxiv.org/abs/2503.02857