Zero-Shot Visual Deepfake Detection: Can AI Predict and Prevent Fake Content Before It’s Created
Ayan Sar ID, Sampurna Roy ID · 2025
Read Zero-Shot Visual Deepfake Detection: Can AI Predict and Prevent Fake Content Before It’s Created by Ayan Sar ID, Sampurna Roy ID (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-18461.txt.
Introduction
Deepfakes threaten trust in media, and detectors usually need examples of each new fake type. Ayan Sar, Sampurna Roy, Tanupriya Choudhury, and Ajith Abraham survey zero-shot detection in this 2025 paper. Zero-shot methods aim to spot fakes they were never trained on. The paper also covers prevention: stopping fakes before they are made.
The survey weighs four detection routes: self-supervised learning, transformer classifiers, generator fingerprints, and meta-learning. Prevention ideas include jamming generator inputs, watermarking real content, live monitoring of creation tools, and blockchain records of origin. The authors list hard open problems: attacks on the detectors, scale limits, ethics, and missing standard benchmarks. They point to future work on explainable detection and mixed media analysis. A joined defense, they argue, needs researchers, security experts, and policymakers together.
Catalog record
| CREATOR | Ayan Sar ID, Sampurna Roy ID |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 181 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-18461