After 15+ years of prosthetics, 3D printing, makeup, glitter and clothing — join the quest to find the most effective deceiver against modern surveillance.
How to read the lenses
Meet the villains.
Smart glasses
Pole-mounted reader
Depth sensor array
2017 — 2020
Depth & near-infrared
Dot projectors and infrared sensors read the face as 3D geometry — in the dark, and with a liveness check a photograph cannot pass.
Adv-Makeup · NIR-LED illumination attacks 2021
IoT home camera
2014 — 2016
CNN face embeddings
DeepFace and FaceNet stop measuring features and start learning them: every face becomes a vector in a space nobody designed by hand.
Adversarial eyeglass frames — Sharif et al. 2016
Dome camera
2010 — 2013
Landmark geometry
Eigenfaces, Fisherfaces, 68-point meshes. The face is a set of distances between named points — and the dome never shows you which way it looks.
URME prosthetic — Leo Selvaggio 2014
Limits: hides the points rather than fooling them. Highly visible to humans.
Bullet CCTV
2001 — 2009
Haar cascades
Viola-Jones slides rectangles of light and dark across the frame. Cheap, fast, everywhere — and it only ever looked for contrast.
CV Dazzle — Adam Harvey 2010
Limits: tuned to the detectors of its day; ineffective against the CNNs that replaced them.
The pyre · pattern recognition, 1960s
Dates mark first public demonstration, not deployment. Every era is a documented result under stated conditions — local, conditional and temporary. Read the archive entries.