Skip to main content

CAPTCHA guide

Arkose Labs / FunCaptcha

Multi-round visual puzzles that get longer as the session looks riskier.

Difficulty: High
Risk engine
Exploded concept drawing of how the Arkose Labs challenge is built

Arkose Labs, once called FunCaptcha, is an enterprise bot-management platform built around adaptive visual challenges that can run over several rounds. Roblox, Microsoft, EA, GitHub, Twitch and many other large sites use it. Difficulty follows risk: a clean visitor may get through with one click, while a suspicious session faces round after round of 3D rotation, shape matching or image selection.

Challenge typesLink to Challenge types

  • 3D object rotation: turn an object until it faces the same way as the target
  • Image/shape matching: choose the picture that matches the reference
  • "Match the pattern": pick tiles in the right sequence
  • Audio / text fallback: rarer, and easier to automate
  • Multi-round enforcement: one challenge after another while the risk stays high

How to identifyLink to How to identify

  • An iframe or script served from arkoselabs.com or funcaptcha.com
  • A data-pkey, public_key or pk= parameter in the widget URL
  • A hidden fc-token input containing a long encrypted string
  • Requests to /fc/api/ or /fc/gt2/
  • Arkose Labs or FunCaptcha branding
  • On enterprise setups, a full-page "Please verify" screen

How it worksLink to How it works

  1. The site loads the Arkose widget with its public key.
  2. Arkose opens a session and serves a challenge, possibly in several rounds.
  3. Once the client solves it, Arkose returns an encrypted fc-token.
  4. The site passes fc-token to its backend, which checks it with Arkose.
  5. The token is bound to the session, the IP and the risk score. Replaying it anywhere else fails.

Free bypass methodsLink to Free bypass methods

Method What it does Success Rate Link
OpenCV + template matching Handles the static image and shape-matching rounds 60-85% opencv/opencv-python
YOLOv8 / CNN classifier Recognises the target object and picks the matching tile 70-90% ultralytics/ultralytics
Angle-regression CNN Predicts the rotation angle in 3D rounds 60-85% custom model (PyTorch/TensorFlow)
Whisper Transcribes the audio fallback when it is available 80-95% openai/whisper
DrissionPage Real Chromium control, well suited to multi-round widgets 60-85% g1879/DrissionPage
Camoufox / Patchright Stealth browsers that keep the risk score down 60-85% daijro/camoufox
undetected-chromedriver Selenium with patches for the widget flow 50-75% ultrafunkamsterdam/undetected-chromedriver

Solving the rotation challengeLink to Solving the rotation challenge

  1. Grab the target picture and the object to turn, from the canvas or an iframe screenshot.
  2. Estimate both orientations. The quick way is keypoint matching (SIFT or ORB) with a homography. The more robust way is a CNN regressor trained on synthetic rotations.
  3. Drag the object along a curved path with variable speed and a slight overshoot.
  4. Let the round's animation finish, and repeat if Arkose serves another round.
  5. Submit the final fc-token.

Practical tipsLink to Practical tips

  • Arkose adapts. A poor fingerprint or a flagged IP brings extra rounds even when every answer is right.
  • Stick to a persistent browser profile and a residential or mobile IP. Datacenter addresses often get stuck in endless enforcement.
  • Never replay an fc-token. Get a new one for each protected action.
  • Rounds are timed. Run the image analysis in parallel, but keep browser actions sequential.
  • When Arkose offers the audio fallback, it is usually the cheapest way through.

Skip the solver: let BrowserSolver handle Arkose Labs

Website AntiBot runs a real browser session with matching TLS, headers, cookies and IP, clears the challenge, and returns the page. You send a URL, you get HTML back.

For education and authorized testing only. Respect the law, privacy, and the terms of the sites you work with. Difficulty labels are editorial estimates. Adapted from the CC0 community field guide on GitHub.