The three apps looked like simple utilities. One identified dog breeds from photos. One identified spiders. One identified insects. Each had between 500,000 and 1 million downloads on Android. Each asked permission to access your camera so it could analyze the photo you pointed it at.
None of them explained that they were uploading your photos — along with your email address, your username, your profile picture, and your GPS coordinates — to a cloud database.
In February, security researchers discovered that all three apps connected to misconfigured Firebase instances with public read and write access enabled. The databases required no authentication to access. Anyone with the URL could read every record. Automated scanners had already indexed them. Researchers found proof-of-concept entries indicating the databases had been accessed by unauthorized parties before the issue was reported.
Approximately 152,000 users’ data was exposed.
What These Apps Were Actually Doing
The apps in question were Dog Breed Identifier Photo Cam, Spider Identifier App by Photo, and Insect Identifier by Photo Cam. Each presents as a simple identification tool: point your phone’s camera at an animal or insect, and the app tells you what it is.
The AI models powering these identifications require backend infrastructure. Running a capable vision model on a smartphone produces good results if the model is large enough and the phone is recent enough. An alternative approach is to upload the photo to a cloud server, run the model there, and return the result.
The apps chose the cloud approach. Photos were sent to servers for processing. That in itself is not unusual or inherently problematic — many mobile apps that use AI features process images server-side.
What the apps also collected, and what they stored in their databases, went beyond what the identification function required.
The databases contained:
Email addresses and usernames. For an app that identifies dog breeds, there is no obvious reason the service needs to know your email address. Accounts linked to real names via email tie the usage to an identifiable person.
Profile photos. A photo of your face stored in a database operated by a dog breed identifier creates a data category — a biometric — that the app’s stated function doesn’t require.
Firebase Cloud Messaging tokens. These are device identifiers used to deliver push notifications. They’re unique to a device and can be used to track device usage patterns across time.
GPS coordinates. The most significant data category. The location of your device at the time you used the app — accurate to a level of precision sufficient to identify your home, workplace, or regular locations.
Why GPS? The most charitable explanation is poor app architecture: collecting location as a default for any app that might conceivably use it, without considering whether this specific app actually needs it. A less charitable interpretation is that location data has commercial value, and the apps were collecting it for purposes beyond their stated function.
The Firebase Misconfiguration Pattern
Firebase is Google’s mobile application development platform. It provides authentication, real-time databases, cloud storage, and hosting for mobile apps. It’s widely used because it’s well-documented, reasonably priced, and enables developers to build cloud-connected apps without managing their own server infrastructure.
Firebase databases require explicit configuration of security rules — policies that define who can read and write what data under which conditions. When a developer doesn’t configure these rules correctly, the default can be a database open to anyone on the internet.
This is not a new problem. Security researchers have been documenting misconfigured Firebase databases for years. The scale of misconfiguration is consistently surprising: tools designed to scan for exposed Firebase instances routinely find thousands of open databases across a wide range of applications.
The pattern in the February 2026 incident is familiar:
- Developer builds an app that needs cloud storage.
- Developer sets up Firebase for convenience.
- Firebase security rules are left at a permissive default or set incorrectly.
- Data accumulates in an open database.
- Automated scanners index the database before developers notice.
- Researchers or malicious actors access the data.
The developers of the three affected apps were apparently unaware of the exposure until researchers contacted them. How long the databases had been accessible is unknown.
Why GPS Data Makes This Worse
Of all the data categories in the exposed databases, GPS coordinates are the most consequential.
An email address exposed in a breach enables phishing. A username enables account correlation across services. A profile photo, if combined with facial recognition, can link an identity to public imagery. These are significant but bounded risks.
GPS coordinates can tell an attacker where you live.
When a person uses an identification app regularly — photographing their dog in the backyard, identifying spiders around their house, checking insects in their garden — the GPS coordinates associated with those uses cluster around their home address. Multiple uses across different sessions create a pattern that resolves to a specific building, not just a city or neighborhood.
For most people, knowing their home address is already public information or easily findable through other means. But for people who have taken steps to keep their address private — domestic violence survivors, witnesses, people with stalkers, journalists and activists working on sensitive topics — precise home location data in a misconfigured database is a direct safety risk.
GPS data also reveals behavioral patterns beyond location. The timestamp of each record indicates when you were at a particular place. A database of usage events over months shows when someone typically leaves home in the morning, when they return in the evening, which locations they visit regularly, and which are occasional. This is pattern-of-life data — the kind of granular behavioral profile that is typically only collected by law enforcement or intelligence agencies with specific legal authority.
A dog breed identification app was building this profile. Passively. Without disclosure. And storing it in an open database.
The Utility App Problem
The three affected apps illustrate a broader problem with how utility apps — apps that serve a simple, practical function — handle the permissions they request.
Camera access is the most fundamental permission for a photo utility app. Without camera access, the app can’t function. Users grant it without hesitation because it’s obviously necessary.
Location access is a different matter. Camera apps frequently request location permission because photos can be geotagged, and geotagging is a useful feature. But geotagging is an opt-in feature in most camera apps. It doesn’t require the app to silently upload your GPS coordinates to a backend database separate from the photo.
The permission model on both Android and iOS has improved in recent years. Both platforms now require apps to justify location access more explicitly, and offer options like “only while using the app” or “precise location” versus “approximate location.” But these controls only work if users engage with them thoughtfully.
When an app asks for location permission with a generic justification — “location helps improve your experience” — there is no disclosure that the specific use is uploading GPS coordinates to a cloud database for retention in a backend that may or may not be properly secured.
The FTC has taken action against apps that collected data beyond their stated purposes. But enforcement is reactive, not preventive. The 152,000 users whose data was exposed in February didn’t have the option of waiting for regulatory action.
How to Approach Photo Utility Apps
The specific apps involved in February’s incident may no longer be in wide use. But the problem they represent — utility apps that access the camera and collect far more than their core function requires — is not specific to three particular apps.
Check permissions before and after installing. On Android, Settings → Apps → [App Name] → Permissions shows what permissions the app currently holds. On iOS, Settings → Privacy & Security → [Permission Category] → [App] shows the same. Review what permissions photo utility apps have, particularly location.
Revoke location from apps that don’t explain why they need it. A dog breed identifier doesn’t need your location. An insect identifier doesn’t need your location. If you granted location permission at install and didn’t notice, you can revoke it. The app may ask again; if it demands location permission to function at all, that’s a red flag.
Prefer apps from known, accountable developers. Apps from developers with established track records, clear privacy policies, and a business model that doesn’t depend on data monetization are lower risk than apps from anonymous developers with no disclosed business model. Free apps with no in-app purchases and no subscription have to generate revenue somehow; data collection is often the answer.
Check privacy policies before granting camera access. This is tedious and most people don’t do it. But for apps with a narrow stated purpose, a privacy policy that describes extensive data collection including location, behavioral analytics, and third-party sharing is a signal worth heeding.
Use on-device alternatives where they exist. For basic identification tasks, on-device models are increasingly capable. Apple’s Visual Look Up can identify plants, animals, and other subjects without uploading photos to a cloud service. Google Lens processes photos for identification purposes through Google’s infrastructure under Google’s privacy policies — which, while not without their own tradeoffs, are at least explicitly stated.
The Larger Point
The three apps that exposed 152,000 users’ data in February weren’t social media platforms with complex data practices. They were simple utility apps that did exactly one thing. And they were secretly accumulating a detailed log of where their users lived and when, stored in an unprotected database.
The assumption that simple apps with limited function are low-risk is not reliable. Camera permission is powerful. Location permission is powerful. The combination of both, in an app that uploads data to an unsecured backend, can create a detailed and sensitive profile without the user ever knowing it’s happening.
Any app that accesses your camera has access to some of the most personal visual data you generate. Where that data goes — whether it stays on your device, goes to a properly secured server, or ends up in an open database — is determined by the developer’s choices and practices, not by the function the app advertises.
For the photos and personal data that matter most to you — memories, documents, personal records — choosing where they live, who can access them, and how they’re secured is worth deliberate decision-making rather than a default grant of permissions to whatever app happens to ask for them.