Harvard students turn Meta's smart glasses into a real-time doxxing tool

Two Harvard students just demonstrated how easily Ray-Ban Meta glasses can turn a walk in the park into an automated doxxing nightmare.

Students AnhPhu Nguyen and Caine Ardayfio built a system called I-XRAY. By chaining together everyday consumer hardware and existing web tools, they created a setup that identifies strangers on the street and pulls up their home addresses, phone numbers, and family members in real time.

An off-the-shelf surveillance rig

The most unsettling part of the demo is that Nguyen and Ardayfio didn’t invent any new surveillance tech. They simply connected dots that already exist.

The project uses the livestreaming feature built into Ray-Ban Meta smart glasses to broadcast video straight to Instagram. A custom computer script monitors the stream, clips faces using computer vision, and runs them through PimEyes, a publicly accessible reverse face search engine.

From there, large language models parse public record databases to correlate identities with real-world information. Within seconds, a mobile app pings the wearer with the target’s full name, phone number, physical address, and even the names of relatives.

In video demonstrations, the creators identified classmates on campus and struck up conversations with complete strangers on public transit, pretending to know them using details harvested on the fly.

Hidden in plain sight

Google Glass notoriously failed a decade ago partly because its clunky, cyborg aesthetic made people deeply uncomfortable. Meta’s latest frames have solved the styling problem, and that is precisely why this demo feels so jarring.

The Ray-Ban Meta frames look identical to standard designer eyewear. While the glasses feature a small LED light to signal when video is recording, real-world testing shows it is nearly impossible to spot in bright sunlight or crowded environments.

Meta’s privacy policy urges users to respect social etiquette and use clear hand gestures or voice commands when recording. When asked for comment, Meta spokesperson Dave Arnold reiterated those exact terms of service. But as this project proves, software guidelines cannot stop someone from quietly weaponizing an inconspicuous camera.

Nguyen and Ardayfio stressed that they have no intention of releasing the code. They built I-XRAY strictly as an awareness project to prove that automated, real-time doxxing is no longer science fiction—it works right now.

Why it matters

For anyone walking through a city, this demo signals the functional death of public anonymity. Even if you avoid social media, any photo of you indexed online can now be matched to your physical person by a stranger wearing standard sunglasses. The authors outlined opt-out guides for databases like PimEyes, but scrubbed records are notoriously hard to maintain across the open web.

For Meta and the wearable industry, I-XRAY exposes a fundamental design conflict. The more natural smart glasses look, the more consumers want them, but that same invisibility strips bystanders of any ability to consent. Relying on polite user etiquette and a tiny indicator light looks increasingly indefensible.

We think the real catalyst here is the addition of LLMs to facial recognition. PimEyes could already find matches, but language models automate the grueling work of cross-referencing public dossiers in seconds. The barrier to instant, real-world stalking has effectively collapsed to zero.