Short answer: yes. Longer answer took me a couple of months and 10 physical devices to confirm.
An Android device leaks somewhere between 20 and 30 identifying attributes to any app that asks — MANUFACTURER, BRAND, MODEL, API level, ABI list, sensor set, GPU vendor string, per-core CPU count, screen DPI, on and on. Individually most of them can be spoofed. Combined, they're a fingerprint that emulators, cloud phones, counterfeit units and dodgy refurbs all struggle to fake cleanly.
The ones that mattered most in practice:
- Emulators — the sensor list is almost always wrong (too few, or all "generic")
- Cloud phones / Docker-based virtual devices — GPU vendor and ABI list rarely match the model they claim
- Counterfeit hardware — CPU core count contradicts the marketing chipset
- Suspicious refurbs — Build.FINGERPRINT is structurally malformed, or the API level doesn't line up with the Android release
I cross-checked roughly 10 real devices against community hardware fingerprint data and the match rate came out better than I'd hoped. Every real device landed inside its own model's fingerprint cluster. The three fakes in the pile — one cloud phone, one BlueStacks instance, one counterfeit "S22 Ultra" — each broke at least three cross-checks.
That result is what became the Hardware Trust Score in DevCheck AI v1.0.4.
The original question
When I first posted this on X, I asked whether anyone actually wanted such a thing before buying a second-hand phone, or whether it was too niche to bother building.
The answer turned out to be a solid yes. v1.0.4 shipped 8 cross-checks — CPU core count, API ↔ Android release, 64-bit ABI consistency, Build.FINGERPRINT structure, SoC ↔ GPU vendor pairing, basic sensor presence, camera presence, ABI ↔ os.arch — collapsed into one 0–100 score with a colour-coded verdict. The features page breaks down what each check looks at and what it costs you when it fails.
