Antidetect browser detection: what these browsers are and how to catch one

How an antidetect browser works

Open one and you get a list of profiles instead of a single browser window. Each profile is a complete stand-in device. A website that asks what it is talking to gets that profile's answers, and nothing ties it to the other profiles on the same laptop. Six parts make that work.

The cheapest spoof needs none of this. Chrome's own DevTools can replace the user agent string and its Client Hints in a few clicks (Chrome docs), and MDN notes that browsers "routinely pretend to be another browser" (MDN). A user agent is a claim a browser makes about itself. Detection starts by testing whether the claim holds.

Who uses antidetect browsers, and are they legal?

Two very different groups buy the same software. Agencies, affiliate marketers and online sellers use it to run many ad, social and store accounts without the platforms linking them; security testers and journalists use it for testing and privacy. Group-IB's list of fraud uses is longer: account takeover, bonus abuse, fake account creation, credential stuffing and synthetic identity fraud (Group-IB).

The browser itself is generally legal. Group-IB's summary: legal for privacy and market research, illegal once it is used for fraud or to get around security systems. Running many accounts can still break a platform's terms where no law is broken. Vendors are open about the multi-account market: Dolphin{anty}'s site lists affiliate marketing, media buying, e-commerce and "bonus hunting" among its use cases (Dolphin{anty}).

In your own app, a fake browser tends to turn up at four moments.

How antidetect browser detection works

A profile is a set of answers the browser gives to JavaScript, the same answers a device fingerprint is built from. Change the answers and the engine underneath stays the same, and so do the graphics chip and the network path. Detection asks two questions: could all these answers come from one real device, and does the engine behave like the browser the answers describe? A good profile gets almost everything right. It only has to get one thing wrong.

On our own test bench we changed one value at a time and watched what Maskbreak's tampering flag reacted to. The bench is a JavaScript spoofing shim, not a commercial browser, but it shows the principle more clearly than a diagram (lab notes).

None of these tells lasts forever. Chrome now ships every two weeks, vendors follow within days or weeks, and each release can remove a contradiction that got noticed. Vendors also go after the classic tells directly: Dolphin{anty} wrote in August 2026 that its engine-level font spoofing stops the real operating system showing through the font list (Dolphin{anty} blog). That is why any catch rate for these tools is a snapshot at best (Maskbreak has not published one), and why a verdict should rest on several independent layers rather than one clever check.

Antidetect browser detection API: what Maskbreak returns

Add the browser snippet to the form you want to protect: signup, login, checkout or a bonus claim. Your server forwards the two tokens it collects to POST /v1/evaluate with your API key and gets one decision back (allow, review or block), with the reasons behind it. In the console the signal is labelled "Fake browser (antidetect)". The API docs list every field; these are the ones that matter here.

The classic setup is a fake browser behind a residential proxy. An example response for that visit, trimmed to the fields that matter:

// example response, POST /v1/evaluate
{
  "decision": "block",
  "risk_score": 85,
  "reasons": ["proxy_detected", "antidetect_browser"],
  "network": {
    "vpn": false,
    "proxy": true,
    "datacenter": false,
    "anonymous": true,
    "tor": false,
    "residential": false,
    "service": null
  },
  "device": {
    "complete": true,
    "checked_signals": ["antidetect", "automation", "emulator", "virtual_machine",
                        "incognito", "privacy_mode", "ip_blocklisted", "high_activity"],
    "antidetect": true,
    "automation": false,
    "virtual_machine": false,
    "tampering_score": 0.0029
  }
}

The score is 85: 50 for the proxy and 35 for the fake-browser flag, and either signal alone would block, so route on decision, not on the number. residential is false because the proxy flag fired; the field means "no proxy or cloud-server flag", not "home connection". service is null because this proxy's operator was not known; when it is, the name appears there. And the tampering score sits near zero while antidetect is true, the pattern we measured on our bench.

No key is needed to see the shape. curl "https://maskbreak.com/v1/evaluate/sample?scenario=bot" returns a synthetic fake, scripted browser on a clean address: block, with antidetect_browser and automation_detected. Its values are a fixture for testing your handler, not a measurement.

Changing the default. The console's Rules tab sets one action per signal for your account, and a rule changes the answer only for the signal it names. So if agencies or sellers are your customers, you can set "Fake browser (antidetect)" to review: a fake browser then goes to review, while a proxy, Tor, a script or an emulator on the same visit still blocks. The other route keeps the block and softens it in your own code where it hurts (review at login, block at a bonus claim). The "VPN + fake browser" rule covers the fake browser only when both fired: set it to review to relax that pair alone, or to block to keep the pair blocked after you relax the Fake browser rule. When a rule changes the answer, engine_decision keeps the default, so you can see what the rule did. Start in watch mode, which logs every decision and lets the visitor through, and switch to enforce once the Events log looks right. The quickstart has the handler, or hand integrate.md to your coding agent.

Time zone, if you forward it. Send the browser's time zone as tz (plain HTTP; the server SDKs do not pass it yet) and a zone from a different country than a clean-looking exit adds timezone_mismatch and moves allow to review. It never blocks on its own, and a careful operator sets the profile's time zone to match the proxy, so a match proves nothing.

What it will not tell you. Which product the visitor used. Whether the account belongs to an agency or a fraud ring: the browser is the same, what the account does next is not. Maskbreak adds evidence to a login; it does not replace passwords, passkeys or MFA. And there is no published, independently benchmarked detection rate for it.

Antidetect browsers in 2026: makers, engines and what changed

The products people search for, with the maker as the vendor's own site names it, the engine it runs on, and what changed recently. Names in blue link to our detection guide for that browser (the Octo Browser one also covers MoreLogin, VMLogin and Hidemyacc); a dash means we could not confirm it from the vendor's site. Most of the larger vendors give away 2 to 5 profiles, so trying one costs nothing. At the top end, Multilogin charges $900 a month for 10,000 profiles billed monthly, or $490 a month billed annually (pricing). Prices change often; the ones below were read on 24 September 2026.

Some names still get searched although the product may be gone:

Proxyway's May 2026 roundup, which uses affiliate links, also covers GeeLark, which sells Android cloud phones, and Nstbrowser (Proxyway). Product names are their owners' trademarks. A listing here says nothing about any vendor's customers or intent; the same software serves agencies and fraud rings. Facts without a link in the row come from each vendor's own site, pricing page or changelog, read on 24 September 2026. The glossary defines the terms used here, from fingerprint to residential proxy.

Antidetect browsers built on Firefox

Most of the market runs on Chromium, the engine behind Chrome. The Firefox-based options as of September 2026: Kameleo's Junglefox, AdsPower's FlowerBrowser (Firefox 153 support since 15 September), the open-source Camoufox, and the Firefox kernels offered by MoreLogin and GPM Login. Dolphin{anty}'s plan table lists a "Chrome or Firefox engine" option, which we could not confirm anywhere else. Multilogin's Stealthfox stays on version 146, and Multilogin recommends its Chrome-based Mimic for new profiles. MASQ was described as Firefox-based in 2024; we could not confirm it is still sold.

For detection, the engine family matters less than whether the profile stays inside it. A Firefox build that claims to be Chrome still behaves and shakes hands like Firefox underneath, which is where a TLS fingerprint can give it away. A Firefox build that claims to be Firefox has to get the version, the platform and the graphics answers right instead, which is where the claim-against-hardware checks come in. Firefox builds also age: a profile stuck on an old engine while claiming a current release is a contradiction of its own.

Questions people ask about antidetect browsers

More on detecting antidetect browsers

Our detection guides for individual browsers are linked in the directory above. These go deeper on the method.