Updated October 2026 · Vaydr Labs · About a 9-minute read

Search for ways to avoid facial recognition and you'll find bold face paint, printed masks, LED hats, patterned hoodies and special glasses. Some come from serious research. Some are art projects. A few are just marketing.

This guide looks at each option the same way: what it actually targets, what published research shows, whether you could wear it on an ordinary Tuesday, and whether it's legal where you live. We make privacy eyewear, so we'll say where glasses fit in. We'll also say where they don't.

The short answer

Method What the research shows Everyday practicality
Bold "dazzle" makeup Designed for an older face detector; its creator has advised against the early looks since 2016 Low. Very noticeable
Algorithm-crafted makeup Strong results in one lab study against one model Low. Needs to be computed for a specific system
Face masks Raised error rates in 2020; newer algorithms closed much of the gap Medium. Restricted in some states
Infrared LED hats Worked in a 2018 research prototype Low. Few tested consumer products
Adversarial clothing Aimed at person detectors, not face recognition Medium. Tuned to specific models
Sunglasses and privacy eyewear Covering the eye region lowers identity matching in human studies; IR-filtering lenses keep the eyes covered for infrared cameras too High. Normal to wear outdoors

No single item makes anyone unidentifiable. The useful question is which layers fit your life.

First: detection and recognition are different

Many "anti-facial-recognition" products target only one step of the process. It helps to know which step.

  1. Detection. The camera software finds a face in the frame.
  2. Analysis. It measures the face: the spacing of the eyes, the brow line, the nose, the jaw.
  3. Matching. It compares those measurements to a database and returns a match with a confidence score.

Some disguises try to stop step 1, so the software never "sees" a face. Others leave detection alone and try to scramble steps 2 and 3. That difference explains why some well-known projects get oversold. Something built to confuse an old face detector says little about a modern matching system.

Many cameras also see more than you do. Security cameras often add infrared light, especially at night. Phone face unlock is the everyday example: Apple says Face ID "captures an infrared image of your face" (Apple). A disguise that only works in visible light may not do much once infrared is involved.

New to how this works? Our walk-through Your day, on camera follows the cameras an ordinary person passes in a single day.

Anti-facial-recognition makeup

The idea: paint patterns on your face that confuse the software.

The best-known example is CV Dazzle, created by artist and researcher Adam Harvey in 2010 as his master's thesis at NYU. It used bold, asymmetric shapes and hair over the face. The original looks were designed against the Viola-Jones face detector, which was widely used at the time. Harvey's own site says that detector has been "unofficially deprecated since around 2016." It also says the project "has advised against using looks 1-5 since 2016," because modern systems use neural networks (Adam Harvey, CV Dazzle). CV Dazzle is better understood as a technique that has to be adapted to each system, not a pattern you can copy.

More recent research is subtler. In 2021, researchers at Ben-Gurion University of the Negev used software to work out where to apply natural-looking makeup so a specific face recognition model (ArcFace) would fail to match. They tested 20 participants on two cameras. Without makeup, the system identified participants in 47.57% of video frames. With random natural makeup, the figure was 33.73%. With the computed makeup, it dropped to 1.22% (Guetta et al., 2021).

Our take: The makeup research is real, but it was computed against one known model in a lab. You won't know which system a store or city camera runs, and bold patterns make you stand out to every person nearby.

Anti-facial-recognition masks and face coverings

The idea: cover the lower half of the face.

The pandemic turned this into a large-scale test. In July 2020, the U.S. National Institute of Standards and Technology (NIST) tested 89 commercial algorithms on photos with digitally added masks. The most accurate algorithms went from failing about 0.3% of the time on uncovered faces to about 5% with masks. Many algorithms failed 20% to 50% of the time (NIST, July 2020).

The industry adapted fast. By December 2020, NIST found that algorithms submitted after the pandemic began had cut masked-face errors by up to a factor of ten. The best made errors on masked faces only "between 2.4 and 5% of the time" (NIST, December 2020). Both NIST studies used digitally applied masks, and results varied widely by algorithm.

There's a second, useful finding. A 2023 study in Scientific Reports asked people to recognize faces wearing either surgical masks or sunglasses. Sunglasses reduced identity recognition more than masks did. The authors concluded the upper face, around the eyes and brows, carries more identity information than the lower face (Or et al., 2023). This was a small study of human observers, not software. Still, it lines up with what algorithms measure: much of a face's identifying detail sits around the eyes.

Our take: Masks still add some friction, but less than in 2020. Wearing one in public is also restricted in some states (see the legal section below).

Anti-facial-recognition hats and infrared LED caps

The idea: hide the face from above, or flood the camera with light people can't see.

A plain brimmed hat or hood is a simple, legal habit. Many cameras are mounted above eye level, and a brim takes away part of their view.

The more technical version uses infrared LEDs. In a 2018 paper, researchers from Fudan University, the Chinese University of Hong Kong and Indiana University showed that infrared light, which people can't see, could be used to fool face recognition. With the LEDs hidden in a cap, they could even get a system to mistake one person for another. They reported over 70% success in finding infrared patterns that could be used this way (Zhou et al., 2018; summary).

Our take: The research is clever, but it's a prototype computed with the researchers' own algorithm. We haven't seen independent tests of consumer "IR hats." A normal cap is still worth wearing.

Anti-surveillance clothing

The idea: wear printed patterns that confuse AI cameras.

The best-known research here is a 2019 University of Maryland and Facebook AI paper, "Making an Invisibility Cloak." It trained patterns on posters and wearable clothing to suppress object detectors, the software that finds people and objects in a scene (Wu et al., 2019). That is a different job from recognizing who a person is. These patterns are also trained against particular detector models.

Our take: Interesting research and good conversation pieces. But a sweater designed to confuse a person detector does little about a camera pointed at your face.

Do anti-facial-recognition glasses work?

Glasses go straight at the part of the face that carries the most identity information: the eyes, the brow and the cheekbones. They're also the only option on this list most people already wear every day.

Two things matter when you compare them.

1. Coverage under infrared, not just how dark the lens looks. A lens's tint tells you how much visible light gets through. Infrared is a different part of the spectrum, and it's what phone face unlock and many night-time security cameras use. A lens can look very dark to you and still look different to an infrared camera.

2. Whether you'll actually wear them. The best privacy layer is the one you have on. A frame that looks like an ordinary pair of good sunglasses gets worn. A novelty disguise stays in the drawer.

Vaydr eyewear uses our GLX lens, which is designed to keep the eye region covered under visible and near-infrared light. That can lower the confidence of a face match. Results vary with the system, the lighting and the angle. A simple way to see the effect is to try face unlock on your own phone while wearing them. Results vary by device. (How the GLX lens works)

  • Recon: a tactical frame for outdoor and everyday wear.
  • Veil: the sport shield.
  • Eclipse: the same features as Veil with a more active look.
Man in a cap and blue shirt wearing Vaydr Recon privacy sunglasses outdoors
A cap and everyday eyewear: two simple layers. Shown: Recon.

From the Lab: For how the GLX lens is specified, see our technical documentation and Vaydr Labs.

Where glasses aren't the answer: Sunglasses don't help with photos already online or with face search engines. They aren't for night driving. And you should take them off whenever an ID check asks you to. That's why the next section is about layers.

Vaydr eyewear is designed to reduce exposure; it cannot guarantee you will not be identified.

A practical, layered approach

No single product handles everything. Start with what's free:

  1. Shrink your photo footprint. Untag yourself, lock down social profiles, and ask friends before posting group photos. Face search engines work from images that are already public.
  2. Use your opt-outs. At many U.S. airport checkpoints, TSA says passengers "may choose not to have a photo taken" and can use their ID document instead (TSA).
  3. Wear a cap or hood when it suits the day. It's simple and legal.
  4. Make eyewear your everyday layer. Choose a frame you'd wear anyway, with lenses that keep your eyes covered under infrared, not only visible light.

Each layer makes you a little harder to identify, and together they add up.

Is it legal to hide your face?

Masks: As of September 2025, the International Center for Not-for-Profit Law counts 23 states plus Washington, D.C. with laws that limit face coverings in public, and notes that "most states do not restrict mask wearing." The laws vary. Some are general bans with exceptions. Others apply only when a mask is worn with a specific intent, such as intimidation, or while committing a crime. Common exceptions include holidays, work, theater and medical reasons (ICNL). Check your own state and city before relying on a mask.

Sunglasses: Wearing sunglasses in public is ordinary. Take them off when asked at ID checks, at security checkpoints, and anywhere staff need to see your face.

Driving at night: Don't wear tinted lenses of any kind after dark or at dusk. Eye-care guidance warns that "ANY tint further reduces the amount of light transmitted to the eye" and impairs night vision (Prevent Blindness).

Never: obscure a license plate, use a signal jammer, or hide your identity to commit a crime. Those are illegal, and they aren't what privacy means.

Frequently asked questions

Does makeup stop facial recognition? Not reliably in everyday life. Classic CV Dazzle looks were built for a face detector that is no longer widely used. Newer algorithm-crafted makeup worked well in a lab against one known model (Guetta et al., 2021), but you won't know which system a camera runs.

Do face masks still confuse facial recognition? Less than they did. NIST found that newer algorithms cut masked-face errors by up to ten times, with the best erring 2.4% to 5% of the time (NIST).

Do infrared LED hats work? A 2018 research prototype with LEDs hidden in a cap fooled face recognition in testing (Zhou et al.). We haven't seen independent testing of consumer IR hats.

Do sunglasses help against facial recognition? Covering the eye region removes much of the face's most identifying detail. In a human study, sunglasses reduced identity recognition more than masks (Or et al., 2023). For infrared cameras, the lens has to keep the eyes covered under infrared light too. Vaydr's GLX lens is designed to do that, which can lower the confidence of a match. Results vary.

Are anti-facial-recognition glasses legal? Sunglasses are legal to wear in public in the U.S. Remove them at ID checks and don't wear tinted lenses for night driving.

What's the single most practical step? Pick layers you'll actually use every day: fewer public photos, your opt-outs, and eyewear you'd wear anyway.

Vaydr is a privacy company for ordinary people. We'd rather you be informed than alarmed. Shop privacy sunglasses →

Sources

0 commentson Anti-Facial-Recognition Makeup, Masks and Clothing: What Works and What Doesn't

Latest Stories

This section doesn’t currently include any content. Add content to this section using the sidebar.