Antidetect Browser Detection
Antidetect browsers, meet layered detection.
One API. Sub-40 ms server-side.
Antidetect browsers — Kameleo, GoLogin, Multilogin, AdsPower, Dolphin{anty}, Incogniton — spin up browser profiles with fabricated fingerprints and pair them with residential proxies to pass as real consumers. Tools that only score the IP see "clean residential ISP" and wave them through. Sentinel evaluates the device, the network, and the device's history with your app on every request — three independent stories that all have to hold up.
KameleoDetected
Desktop antidetect with custom Chromium and Firefox engines, locally stored profiles, and Selenium / Puppeteer / Playwright hooks. How the tampering score, automation signal, and proxy flags stack up against it.
How we flag Kameleo →
GoLoginDetected
Cloud-stored profiles on the Orbita engine, team sharing, a free plan, and a run-in-the-cloud mode. Why cloud profile farms surface in device history and network signals first.
How we flag GoLogin →
MultiloginDetected
The longest-running commercial antidetect — Mimic (Chromium) and Stealthfox (Firefox) engines, team workspaces, automation APIs. What two custom engines cost in consistency.
How we flag Multilogin →
AdsPowerDetected
Volume-priced antidetect popular in cross-border e-commerce, with no-code RPA automation and bulk profile tooling. Where the automation and device-reuse signals bite hardest.
How we flag AdsPower →
Dolphin{anty}Detected
Team-oriented antidetect from the affiliate and media-buying world — profile tags and statuses, bulk proxies, scriptable scenarios. How scenario automation and reuse history expose it.
How we flag Dolphin{anty} →
IncognitonDetected
Budget antidetect with a ten-profile free tier, bulk profile creation, and Selenium / Puppeteer integration. What high-volume, lower-effort profiles trip first.
How we flag Incogniton →
How antidetect detection actually works
Every antidetect browser is a rebuilt Chromium or Firefox whose fingerprint surface — the properties websites can read about a device — has been made configurable. The point is to make each "profile" pass as a different real device. Sentinel's device layer answers two questions about every session: does this browser carry the traits of that class of tooling (the antidetect-browser signal), and how heavily has the fingerprint surface been altered (the browser-tampering score, 0–1, where anything above 0.6 strongly suggests an antidetect browser)?
The tampering score works because faking a whole device consistently is genuinely hard. The claimed browser version has to match how the engine actually behaves; the claimed operating system has to match the platform internals; every spoofed surface has to agree with every other one. Contradictions push the score up — and the device is only one of three stories a session tells; the network and the history are checked independently.
Antidetect-browser signal
Fires on the category of tooling — a fingerprint surface rebuilt to be configurable — not on any one vendor's build, so new releases don't reset detection. Returned as device.antidetect.
Browser-tampering score
device.tampering_score, 0–1. Above 0.6 is strong evidence of an antidetect browser; 0.3–0.6 flags softer engine/OS inconsistencies; below 0.3 is normal traffic.
VM & emulator detection
Profile farms run seats in virtual machines and emulators. The host environment is scored independently of the browser profile — device.virtual_machine, device.emulator.
Bot & automation signal
Every major antidetect advertises Puppeteer, Selenium, or Playwright integration. Machine-driven profiles trip device.automation with the automation_detected reason code.
The network tell
Antidetect users ride residential proxies and VPNs — Sentinel's Spur-powered network layer flags those exits on its own evidence, no matter how clean the device looks.
Device history
device.times_seen counts a device's visits to your app, and multi-account linking (device.linked_accounts) exposes one machine behind many "different" customers — the farm-reuse tell.
One honest caveat belongs on this page: no fraud vendor catches every antidetect session every time — these tools are commercial products that update specifically to defeat detection. That's why Sentinel returns each signal separately alongside a 0–100 risk score instead of a bare yes/no: layered signals plus your own rules produce the verdict, so one patched tell never blinds the pipeline. The per-tool pages above cover each product and the signals its sessions run into; full response fields for device.antidetect and device.tampering_score are in the API docs. Tools without a page of their own — Octo Browser, MoreLogin, VMLogin, Hidemyacc — share the same structural weakness, covered in the second tier of antidetect tools; Linken Sphere has its own breakdown.
One POST. Sub-40 ms server-side. Free during open beta.
Free tier: 1,000 requests/hour. No credit card, no expiry. Sign up in minutes — or scan your own browser first.