VerbaFake, EchoFake, and PixelFake: lightweight architectures with advanced augmentation for unimodal deepfake detection
Usman Ali, Muhammad Rizwan Shoukat, Mahjbeen Rajput, Saeed Ali Omar Bahaj, Muhammad I Khan, Tariq Mahmood · 2026
Read VerbaFake, EchoFake, and PixelFake: lightweight architectures with advanced augmentation for unimodal deepfake detection by Usman Ali, Muhammad Rizwan Shoukat, Mahjbeen Rajput, Saeed Ali Omar Bahaj, Muhammad I Khan, Tariq Mahmood (2026) 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/pmc-13500687.txt.
Description
Ali and colleagues built lightweight detectors for fake text, audio, and images, three separate problems gathered under the deepfake label. Most detectors are heavy models that smaller newsrooms and police units cannot run quickly. Their 2026 framework trains one specialist pipeline per format, using heavy data augmentation to harden the models. It includes three home-grown compact networks named VerbaFake for text, EchoFake for audio, and PixelFake for images, alongside well-known efficient architectures.
Trained on tens of thousands of samples per format, the text models scored near-perfect accuracy in the reported tests. The best image models reached the high nineties, and the audio trio climbed just below that mark. The authors' scratch-built audio model beat two famous imported networks in their comparison. Because each detector works on one format alone, the system stays small and fast enough for real-time screening.
Catalog record
| CREATOR | Usman Ali, Muhammad Rizwan Shoukat, Mahjbeen Rajput, Saeed Ali Omar Bahaj, Muhammad I Khan, Tariq Mahmood |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |