security

9 Million Face Photos Leaked From a Service You Never Joined

A reverse-image lookup site left 9 million face photos in an open AWS bucket. Here's how your face ended up in their database without your knowledge.

Security researcher Jeremiah Fowler found an unsecured Amazon S3 bucket earlier this month belonging to ClarityCheck, a reverse-image lookup service that advertises itself as “private and secure.” The bucket held roughly 450 GB of data — including more than 9 million image files organized into folders named faces and profiles.

The images included photos of adults, teenagers, and children. The database also contained email addresses and phone numbers. Anyone who knew where to look could access all of it through a URL embedded in the company’s own public website code.

The unsecured database has since been locked down. But the breach raises a question most people have never thought to ask: how did your face end up in a database belonging to a company you’ve never heard of?


What ClarityCheck Actually Does

ClarityCheck describes its product as a way to “use reverse image search to identify anyone in a photo. Find names, social profiles, and online presence in seconds.”

The product pitch sounds like a privacy tool — search for your own face to find where it appears online, then request takedowns. But the service is unverified in both directions. Anyone can upload any face. There is no requirement that you’re searching for your own image. And the platform returns results linking to pages across the web, giving users a map of someone’s online presence from a single photo.

To build that search capability, ClarityCheck — like similar services — has to crawl the web and index face images. It doesn’t ask permission. It doesn’t notify people whose faces appear in the results. If your face appears on a publicly accessible webpage, it may have been scraped, processed into a biometric faceprint, and added to the database without your knowledge.

This is the architecture behind every consumer face-search service. The difference with ClarityCheck is that an unsecured S3 bucket made the entire backend — the raw photos, the profiles, the contact data — accessible to anyone online.


How Faces Get Into These Databases

You don’t have to sign up for a people-search service to end up in one.

Consumer face-search engines build their databases by crawling publicly accessible web pages and extracting face images. Sources include:

  • Social media profiles, even ones with limited visibility settings
  • Personal websites, blogs, and online portfolios
  • News articles and publication photos
  • Professional directories and LinkedIn profiles
  • Photos attached to forum accounts, reviews, and comment sections
  • Wedding websites and event photo galleries

The photos are converted into biometric faceprints — mathematical representations of facial geometry — and stored for search. The original image, the source URL, and often associated metadata (names, usernames, page titles) are retained.

The scale is significant. ClarityCheck’s exposed database held more than 9 million images. PimEyes, a larger competitor, claims to index roughly 3 billion face images. These are not small datasets compiled from formal records. They’re scraped from the public web at a scale that makes opting out impractical without removing yourself from the internet entirely.


What the Breach Exposed

The specific exposure in the ClarityCheck case extended beyond face images to operational data:

Email addresses and phone numbers — linked to user accounts, not to the people in the photos. This means ClarityCheck’s own customers are exposed in addition to the subjects of the images.

Photos of minors — the database included images of teenagers and children, drawn from the same web-crawling process used for adults. There are no age filters in how this content was collected.

File structure revealing operational intent — the folder names faces and profiles in an exposed database are not incidental. They suggest the company was building structured biometric profiles, not just archiving random images.

A gap between marketing and reality — ClarityCheck explicitly marketed its service as “private and secure.” An 450 GB open S3 bucket is neither.


Why This Matters Beyond One Company

ClarityCheck is one company with a particular carelessness about security. But the problem this breach illustrates applies broadly.

Reverse-lookup services are structurally privacy-invasive. The product only works if the database contains faces that people didn’t knowingly submit. The entire value proposition depends on crawling and indexing without consent.

Cloud storage misconfiguration is common. Thousands of S3 buckets with sensitive data have been found exposed over the past several years. The underlying storage infrastructure is secure if configured correctly — but “configured correctly” is the constraint that fails repeatedly. A company holding millions of biometric images has taken on a liability it may not have the competence to manage.

Biometric data is not reversible. If your face is in a database and that database is breached, you cannot change your face the way you can change a password or a credit card number. The exposure is permanent. Every face-search engine that has indexed your image is a new point of failure.

Legal protection varies dramatically. Illinois’s Biometric Information Privacy Act (BIPA) requires explicit consent before collecting biometric data — and has produced major enforcement actions. Most other US states have no equivalent protection. In the EU, biometric data is classified as a special category under GDPR, requiring explicit consent for processing. But many face-search services operate from jurisdictions with weaker rules.


What the Researchers Found in the Code

Fowler’s discovery of the open bucket wasn’t accidental. The URL exposing the bucket was embedded in ClarityCheck’s own publicly available website code — essentially hiding in plain sight for anyone who viewed the site’s source.

This points to a common failure pattern. Cloud storage buckets default to private, but misconfiguration — enabling public access, embedding credentials or URLs in client-side code, setting overly broad bucket policies — is how they end up exposed. The failure isn’t exotic. It’s the product of moving fast without auditing the result.

For a service holding biometric data and contact information on millions of people, that kind of operational carelessness is not a minor gap. It’s a fundamental mismatch between the data being held and the security applied to it.


What You Can Do — and What You Can’t

The honest answer is that your options are limited.

Search for your own face. Services like PimEyes, Epieos, and similar tools let you upload a photo and see where your face appears online. This can identify unexpected sources — event photos, old forum profiles, news articles — that you might then request be taken down or de-indexed. It’s incomplete, because not every service shows you everything it holds, but it’s a start.

Reduce your publicly accessible photos. The fewer publicly crawlable images of your face, the smaller the footprint. This means reviewing what’s visible on social media, professional directories, and any public-facing profiles. Many social platforms let you restrict photo visibility to connections only.

Submit opt-out requests to people-search services. ClarityCheck, PimEyes, and similar services have formal opt-out mechanisms. The quality of opt-out compliance varies. These are not permanent — databases are refreshed and your data may be re-added after some time.

Understand what’s beyond your control. If a third party — a photographer, a news outlet, a wedding website — publishes a photo of you on a publicly accessible page, it’s likely to be indexed by face-search services unless you proactively request removal. You cannot prevent indexing in advance.


The Broader Question About Who Holds Your Images

The ClarityCheck breach makes tangible something that’s otherwise abstract: the images and data you’ve distributed across the web aren’t under your control once they’re out.

Every platform, every service, every photo-hosting site that holds images of you is a potential point of exposure. The question isn’t just whether a company advertises itself as secure. It’s whether their operational practices — their infrastructure, their access controls, their storage configuration — actually match that description.

A company that stores millions of biometric images in a cloud bucket that can be accessed by anyone with the right URL has made a choice. The marketing said “private and secure.” The S3 configuration said otherwise.

What you store, and where, matters. The photos you keep in private, access-controlled storage are not the ones that end up in a researcher’s breach disclosure. The ones that end up in breach disclosures are the ones left somewhere public by a company whose security practices didn’t match their privacy claims.

Your memories deserve better than an ad platform.

Try daftei free →
← All posts