AntiFake
AntiFake detects forged face videos in real time. It was developed under the Ministry of SMEs and Startups Startup Growth Technology Development Project, and an external testing institute measured it on eight items.
중소벤처기업부 창업성장기술개발(R&D) · Period 2023.05 – 2024.07
AI technology that detects forged face videos in real time,
developed through a national R&D project.
As more processes verify identity by face, such as remote account opening, online identity verification and video interviews, technology is needed to confirm that the face is real.
Technology that analyzes forgery after the fact already exists. This project aimed at real-time detection, delivering a verdict while the verification is still in progress.
Performance evaluation results from an external testing institute.
The Gumi Electronics and Information Technology Research Institute (GERI) measured eight items, and every item exceeded its target.
| Test item | Unit | Target | Measured | Verdict |
|---|---|---|---|---|
| Desktop measurements GPU RTX 3090 · CPU Intel Core i9-10900F · RAM 32GB | ||||
| Face image feature point extraction accuracy | % | 95 or higher | 99 | PASS |
| AntiDeepFake face extraction accuracy | % | 93 or higher | 99.84 | PASS |
| Face image feature point extraction time | ms | 33.3 or lower | 2.043 | PASS |
| AntiDeepFake face extraction time | ms | 222 or lower | 7.283 | PASS |
| Mobile measurements Samsung Galaxy S21 Ultra (Android) · iPhone 13 Pro Max (iOS) · 5G | ||||
| Face image feature point extraction speed on mobile | FPS | 20 or higher | 70.3 Android 74.3 iOS | PASS |
| AntiDeepFake face extraction speed on mobile | FPS | 7 or higher | 59.2 Android 41.8 iOS | PASS |
| Processor usage during feature point extraction on mobile | % | 60 or lower | 46.8 Android 40.8 iOS | PASS |
| Processor usage during AntiDeepFake detection on mobile | % | 90 or lower | 48.2 Android 66.68 iOS | PASS |
- Testing institute
- 구미전자정보기술원(GERI)
- Test period
- 2024.03.29 – 2024.04.18
- Issue date
- 2024.04.26
- Test location
- 주식회사 브이로프
- Dataset
- FaceForensics++ (DeepFakes · Face2Face · FaceSwap · NeuralTextures)
- Method
- Average of 5 repeated measurements on 30 samples
AntiFake can be applied in the following areas.
-
eKYC · Online identity verification
Determines whether the face video submitted during account opening or identity checks is genuine footage.
-
Remote identity verification
Confirms that the other party is being filmed live during identity verification over video.
-
Video evidence verification
Applied to procedures that must determine whether submitted video has been manipulated.
-
Content platform filtering
Applied to services that must filter out manipulated videos at upload.
Free to download on the App Store.
| Released | 2024.02.06 |
|---|---|
| Supported languages | Korean · English · Traditional Chinese · Japanese |
| Availability | Downloaded in 25 countries (App Store, as of 2024.09.25) |
| Download | App Store |
Integration with enterprise systems, technology transfer and licensing are discussed separately. Tell us about your environment and we will explain the available options.
Carried out as a national R&D project.
| Project title | AI 적용, 얼굴 위변조 영상 실시간 탐지기술 개발 |
|---|---|
| Program | 중소벤처기업부 창업성장기술개발(R&D) |
| Managing agency | 중소기업기술정보진흥원(TIPA) |
| Project period | 2023.05 – 2024.07 |
| Principal investigator | (CEO) |
| Total R&D budget | KRW 150 million (government grant KRW 120 million) |
| Deliverable | AntiFake (software) |
This technology is the result of the Startup Growth Technology Development Project (창업성장기술개발사업, Didimdol track) conducted by the Ministry of SMEs and Startups (중소벤처기업부).
Related patents.
The CEO is the inventor on all three, and VROF uses them under a patent license agreement with its affiliate.
| Type | Title · number | Filed | Registered | Holder · inventor |
|---|---|---|---|---|
| Registered | System and Method for Producing Special Effects through Face Recognition 10-2529209 | 2021.09.16 | 2023.04.28 | 주식회사 이엔터 Inventor et al. |
| Registered | Distributed Processing System for Real-time Deepfake Image Detection in Mobile Environments 10-2896820 | 2022.12.28 | 2025.12.02 | 주식회사 이엔터 Inventor et al. |
| Filed | AfterRemaker (Class 09) 10-2024-0022442 | 2024.02.16 | Under examination | 주식회사 브이로프 Inventor et al. |
VROF designs its collectors to handle interruptions.
Collection jobs that run for days can be interrupted by slow responses, format errors or computer shutdowns. Our collectors are built from the start to resume where they stopped and refill only the missing parts, so you receive complete data.
-
01
Reads the values the page already holds, not the screen
Selecting elements by screen structure means a site redesign can quietly pull in the wrong values.
-
02
Remembers where it stopped
Large-scale collection is always interrupted at some point. On restart, it resumes from the saved progress point.
-
03
Re-collecting never overwrites earlier data
Records are merged by unique identifier. Data already collected stays intact, and only duplicates are removed.
-
04
Failed items are logged and re-collected on their own
URLs that fail are recorded separately, and a script re-collects only that list.
-
05
Switches routes when blocked
If a response does not arrive or comes back malformed, it switches to another route and continues from that point.
-
06
Delivered ready to run without a development environment
Delivered as an executable so staff can use it without installing Python.
Our longest-running collection system covered over 40 sites for 5 years, maintained continuously as target sites changed their structure or blocked access.
Multiple servers, one GPU cluster
AI models too large for a single server are split across several GPU servers and run together. Inferaft, a service we operate ourselves, runs on this setup.
- Connected serversSeveral GPU servers are networked and used as one.
- Distributed processingLarge models and high request volumes are spread across the GPUs on each server.
- In productionInferaft handles live subscription traffic at inferaft.com.
Technologies used over 23 years of development.
The CEO designs and builds each project. Depending on project size, we select and bring in the developers needed. We do not pass requirements on to subcontractors.
- Python
- PHP
- Linux
- HTML
- Docker
- AWS
- EC2
- GitHub
- Redis
- MongoDB
- Node.js
Three full-time staff, with no outside subcontracting.
Talk to VROF about adopting AntiFake or technical partnerships.
Tell us about your environment and what you need to detect, and we will review how it can be adopted.