Contest the spreading of biometric surveillance.
Ghostmaxxing is a public research lab for testing face-recognition camouflage in the browser, documenting what actually works, and investigating the biometric systems already installed in public space.
It is a test bench, an archive of what has already been tried, and a way for people who know how these systems are (ab)used, to reach us.
Why Ghostmaxxing?
Face recognition arrived in public space pushed by the idea that it is an advantage, for collective security and for personal benefit. That is already debatable, and the evidence of abuse keeps piling up, but meanwhile it is getting normalised through consumer products.
The challenge of Ghostmaxxing is socio-politica! Face-recognition abuses come the collective inability to challenge the techno-securitarian story we keep getting sold. It is not just software what's deploy in cameras, and so is not just software what we need to change it. Ghostmaxxing wants to be a method: to raise awareness, talk to new audiences, and work globally.
Dreamy Vision.
- Make adversarial makeup a popular practice.
- Make workshops easier to run.
- Support and test any forms: clothing, 3D-printed objects, hats, lights and beyond.
- Make a worldwide scoreboard of the most successful and subtle tactics.
- Find the weak spots in every face-recognition system.
Who is behind it.
Ghostmaxxing is developed by Claudio Agosti. Started for the 2026 NINA festival, where it ran as Ghòstati (become a ghost), a workshop held with the participation of Michelle Tylicki. (Michelle had previous experience in adversarial makeup, and has hosted the installation DAZZLE).
It started as a tool to support experimental workshops, but then I reframed it as Ghostmaxxing in an attempt to make the personal experiment part of a collective one. I added the submission platform, the Fediverse integration, the cultural references, and complementary projects selection.
Ghostmaxxing has now the ambition of making it a worldwide
archive of effective camouflage, be an hub to find new technology, testing them, and report if and how they work.
So it became a standalone project, though the whistleblowing processing and workshop execution are organized by NINA.
How Does Ghostmaxxing Work?
Play with Faces
With water-based makeup, with privacy-clotches brands,
or anything else worth trying. A result can be shared,
so others can try to replicate and check if works for them too!
Every result is local: you share it only if you want to.
Check out the lab
Open Ears
This lab uses open-source components, and it alone cannot tell us if the techniques works out there. But, if you build, sell, operate, encounter face recognition, you know it! And if you don't? suggest to who might do it to:
Leak to us safely
Federated Mouth.
Let's find the least invasive method to break face recognition. A federated social network is the natural environment for sharing results, inviting replications and learning more. Everything this project touches gets reported there. The projects is natively integrated in the Fediverse:
Follow it there
Wouldn't be needed if wasn't a "ban" that keeps getting reopened.
In Europe, real-time remote biometric identification in publicly accessible spaces for law enforcement is treated as a prohibited AI practice, subject to narrow exceptions. The Reclaim Your Face campaign spent years getting that prohibition written, but none of this matters to the surveillance state. It starts with "special" cases, then pilot deployments, then the Data Protection Authority blocks them, and then the surveillance state cripples the Data Protection Authority. It would not be a fight if it were that easy.
What we need: informants!
The reporting node runs on GlobaLeaks, and there is a dedicated questionnaire for each kind of informant.
Read the safety guidance on the reporting page before you write anything. Describe what you saw in your own words rather than uploading originals where you can, and do not continue if doing so would put you at risk.
Fork the code.
- ghostmaxxing (on GitHub)
- This site and the browser lab. It runs locally: clone it, serve it, and the whole test bench works on your own machine with no server involved. The documentation is fairly extensive. Take a look.