Every build is checked by seven detectors, with dated results and history kept in the product. This guide explains what each test measures, how to read the verdicts and why one requires manual review.
The latest run on engine 154.0.8037.58 scored 100 out of 100, with responses from all six weighted detectors. Here is what each one measures:
| Detector | What it measures | Latest result |
|---|---|---|
| CreepJS | fingerprint coherence and conflicting signals | ok · 0% headless, 0% stealth |
| iphey | browser, hardware and network integrity | Trustworthy |
| bot.sannysoft | automation, WebDriver and headless indicators | 0 failures, WebDriver absent |
| Cloudflare | access through a live anti-bot challenge | challenge passed |
| TLS | JA3/JA4 TLS handshake fingerprint | JA4 t13d1517h2, Chrome family |
| WebRTC | real IP address leakage | no leak detected |
| pixelscan | screen, location and network coherence | reviewed manually |
A TLS fingerprint is captured before a page loads. The order of ciphers and extensions in the handshake produces JA3 and JA4 hashes. These depend on the engine's network stack, so a browser that changes only its User-Agent is exposed on the first connection. On Chromium 154, our engine produces t13d1517h2 , the expected value for this Chrome version. For comparison, our Firefox engine produces t13d1617h2 and a User-Agent containing rv:152.0: a coherent Firefox identity, not Chrome with a different label.
Pixelscan does not expose its final result as page text. After a minute, automated text extraction sees only navigation and ad blocks, where the word automation appears nine times, making keyword-based scoring unreliable.
manual: NimbusX stores a screenshot, excludes the test from the combined score and leaves the verdict to a reviewer. One invented number would undermine every measured result.Every run is stored with its date, profile ID and result, including failures. A red row is often more useful than a green one: it shows which build failed which detector and what changed after a Chromium update.
One example comes from our own history. After we added device configurations, the number of highlighted CreepJS sections fell from two to one, and the test runner recorded the improvement automatically. An iphey tab that fails to load is treated differently: the page can also crash on an unmodified engine, so a detector with no response is excluded rather than counted as a pass or failure. Regression data is useful only when its rules are trustworthy.
Without installing NimbusX fingerprint test measures your current browser. In the installed product, the test bench opens selected detectors in fresh profiles and stores each site’s actual response:
How fingerprint checks connect profiles · Proxies that cause the browser to hang