CCDF
CCtv DeepFakes
A video deepfake detection benchmark with 1,840 surveillance videos (460 real, 1,380 generated) spanning 16 crime/accident categories.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info ccdf
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
10 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| B-Free | 0.6 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| NSG-VD | 0.6 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| STALL | 0.6 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| D3-Vid | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| WaveRep | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| D3-Im | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| ForgeLens | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| C2P-CLIP | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| C-Forensics | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
| L3DE | 0.5 | N/A | Independent | medium | Oct 6, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
CCDF contains realistic surveillance footage of crimes and accidents, with synthetic videos generated by Grok Imagine, VEO 3.1, and Sora 2. It offers raw, cleaned (standardized encoding), and altered versions to prevent shortcut learning.
Evaluated against ten state-of-the-art detectors, none reliably distinguished generated from real videos, exposing gaps in current evaluation datasets despite strong performance on existing benchmarks.