SynthForensics: Benchmarking and Evaluating People-Centric Synthetic Video Deepfakes
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913103396995072 |
|---|---|
| author | Leotta, Roberto Sambataro, Salvatore Alfio Ragaglia, Claudio Vittorio Casu, Mirko Petralia, Yuri Guarnera, Francesco Guarnera, Luca Battiato, Sebastiano |
| author_facet | Leotta, Roberto Sambataro, Salvatore Alfio Ragaglia, Claudio Vittorio Casu, Mirko Petralia, Yuri Guarnera, Francesco Guarnera, Luca Battiato, Sebastiano |
| contents | Modern T2V/I2V generators synthesize people increasingly hard to distinguish from authentic footage, while current evaluation suites lag: legacy benchmarks target manipulation-based forgeries, and recent synthetic-video benchmarks prioritize scale over realistic human depiction. We introduce SynthForensics, a people-centric benchmark of $20{,}445$ videos from 8 T2V and 7 I2V open-source generators, paired-source from FF++/DFD reals, two-stage human-validated, in four compression versions with full metadata. In our paired-comparison human study, raters prefer SynthForensics in $71$--$77\%$ of head-to-head comparisons against each of nine existing synthetic-video benchmarks, while facial-quality metrics fall within the FF++/DFD baseline range. Across 15 detectors and three protocols, face-based methods drop $13$--$55$ AUC points (mean $27$) from FF++ to SynthForensics and a further $23$ under aggressive compression; fine-tuning closes the gap at a backward cost on legacy benchmarks; training from scratch shows synthetic and manipulation features largely disjoint for most detectors. We release dataset, pipeline, and code. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_04939 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SynthForensics: Benchmarking and Evaluating People-Centric Synthetic Video Deepfakes Leotta, Roberto Sambataro, Salvatore Alfio Ragaglia, Claudio Vittorio Casu, Mirko Petralia, Yuri Guarnera, Francesco Guarnera, Luca Battiato, Sebastiano Computer Vision and Pattern Recognition Modern T2V/I2V generators synthesize people increasingly hard to distinguish from authentic footage, while current evaluation suites lag: legacy benchmarks target manipulation-based forgeries, and recent synthetic-video benchmarks prioritize scale over realistic human depiction. We introduce SynthForensics, a people-centric benchmark of $20{,}445$ videos from 8 T2V and 7 I2V open-source generators, paired-source from FF++/DFD reals, two-stage human-validated, in four compression versions with full metadata. In our paired-comparison human study, raters prefer SynthForensics in $71$--$77\%$ of head-to-head comparisons against each of nine existing synthetic-video benchmarks, while facial-quality metrics fall within the FF++/DFD baseline range. Across 15 detectors and three protocols, face-based methods drop $13$--$55$ AUC points (mean $27$) from FF++ to SynthForensics and a further $23$ under aggressive compression; fine-tuning closes the gap at a backward cost on legacy benchmarks; training from scratch shows synthetic and manipulation features largely disjoint for most detectors. We release dataset, pipeline, and code. |
| title | SynthForensics: Benchmarking and Evaluating People-Centric Synthetic Video Deepfakes |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.04939 |