A device farm is many devices that are suspiciously alike, arriving from one network, doing the same thing. No single device can see that about itself. Fingerly combines what a device reports with what your traffic shows across devices.
One device looks innocent. A rack of them does not. Device farm detection has two halves. A native app can tell when its own device looks provisioned for bulk use rather than owned by a person. The server sees the other half: many near-identical devices behind one network in a short window.
Default weights per platform. When both halves of a device farm fire, confidence rises and the weight is still counted once.
| Signal | Web | Android | iOS | What it means |
|---|---|---|---|---|
| Device farm | 7 | 8 | 8 | The device looks provisioned for bulk use, or many alike devices share its network. |
| Android emulator | n/a | 9 | n/a | The app runs on an emulator or a virtualised Android. |
| iOS simulator | n/a | n/a | 16 | The app runs in the iOS Simulator. |
| Cloned app | n/a | 9 | n/a | The app runs inside a cloning framework, next to copies of itself. |
| Virtual machine | 14 | n/a | n/a | The browser runs in a virtual machine. |
| High activity | 6 | 5 | 6 | One device is far busier than your traffic normally is. |
The cross-device half is learned from your traffic, so what counts as unusual is set by your own product.
Native SDKs report when a device looks like it came out of a box, was imaged and never used by a person.
The server counts distinct devices of the same kind behind one network over five minutes, one hour and one day.
Your own traffic sets the line. At least five distinct devices are needed, and the detector learns for seven days first.
Either half fires the signal. Both together raise its confidence, and the weight is counted once.
Households, shops and schools share devices and networks. The signals are built so that ordinary sharing does not look like a farm.
Two accounts on one device is ordinary. Treat a device farm finding as a reason to review, and let the cluster size guide the action.
Phones rented in a datacentre are not emulators. They are caught by the device farm half, which asks whether anyone has used the device.
Managed work profiles are a normal Android setup, and are weighted so that an enterprise fleet is not reported as cloned apps.
Device farms exist to turn one operator into many customers. Check at the moments that reward being many.
A hundred welcome bonuses, one operator.
Identify at signup and again when a reward is claimed. Review the clusters, not every customer.
Signups from devices nobody owns.
Emulators and provisioned devices creating accounts are a strong reason to hold verification until something more is known.
Rewards farmed before they launch.
Check eligibility before an airdrop or a referral payout, while the connected accounts can still be reviewed together.
Only partly. A browser cannot see most of what makes a phone look provisioned for bulk use, so on the web the signal leans on the cross-device half: many alike devices behind one network. Native iOS and Android apps report both halves.
The cross-device half compares activity with your own traffic. It learns for seven days and needs at least 1,000 samples before it adds to the score, so a launch week does not look like a farm.
No. Developers and testers use them every day. That is why emulators and simulators are signals with weights you control, and why development and staging keys exist for your own testing.
Start with the signals, keep your own decisions, and pay only for what you identify. No credit card needed.