Skip to main content

CAPTCHA guide

Binance CAPTCHA

Binance's own slider puzzle, tied to a device fingerprint and a signed request.

Difficulty: High
Risk engineImage puzzle
Exploded concept drawing of how the Binance challenge is built

Binance runs its own CAPTCHA stack, bCAPTCHA2, on login, sign-up, password recovery and other sensitive actions. The first challenge is normally a slider puzzle, and repeated failures can turn it into a grid or box selection. The answer is tied to a device fingerprint and a signed request, so dragging the piece to the right spot is not enough on its own.

How to identifyLink to How to identify

  • Traffic to accounts.binance.com and bin.bnbstatic.com
  • POST /bapi/accounts/v1/public/account/security/request/precheck answering with captchaType: "bCAPTCHA2"
  • POST /bapi/composite/v1/public/antibot/getCaptcha
  • Response fields sig, salt, path2, ek, captchaType and tag
  • POST /bapi/composite/v1/public/antibot/validateCaptcha
  • A token beginning with captcha#
  • Headers such as device-info, captcha-sdk-version, x-captcha-se and bnc-uuid
  • The slider label "Slide to complete the puzzle"

How it worksLink to How it works

  1. The client calls the precheck endpoint with an email and business type, and gets back captchaType, sessionId, validateId and validationTypes.
  2. getCaptcha returns the challenge details: sig, salt, path2 (a relative image path), ek, captchaType and tag.
  3. The image is fetched from https://bin.bnbstatic.com followed by path2.
  4. After the slide, the client sends the encrypted mouse path and fingerprint as data, with a signature s, to validateCaptcha.
  5. On success it receives a captcha# token, which is submitted with sessionId and validateCodeType.
  6. If the risk score stays high, the challenge escalates to a BOX grid selection.
  7. None of this is documented and it changes over time. Treat field names as unstable.

Free & open-source bypass methodsLink to Free & open-source bypass methods

Tool What it does Success Rate Link
PuzzleSolver Detects the gap and drags like a person 70–90% vsmutok/Puzzle-Captcha-Solver
OpenCV matchTemplate Canny edges plus template matching to find the gap 75–92% opencv/opencv-python
DrissionPage / Playwright Drives the real widget so the trajectory is sent in-session 60–85% g1879/DrissionPage · microsoft/playwright-python
Patchright / Camoufox Cuts down fingerprint and automation signals 50–80% Kaliiiiiiiiii-Vinyzu/patchright-python · daijro/camoufox
YOLOv8 Finds the targets in the BOX grid fallback 80–95% (trained) ultralytics/ultralytics

Practical approachLink to Practical approach

  1. Work in a real browser with a stable device fingerprint and a residential IP.
  2. Reach the challenge through the normal login or precheck flow. Calling the hidden endpoints directly is a red flag.
  3. Take the original background and piece images from the getCaptcha response.
  4. Find the gap with Canny + matchTemplate, correcting for the displayed image scale.
  5. Drag with acceleration, a slight overshoot and some jitter.
  6. Let the page send its encrypted trajectory as usual. Swapping it for a bare x-coordinate will fail.
  7. Keep the captcha# prefix and submit the token with its matching sessionId.

Practical tipsLink to Practical tips

  • path2 is relative, so prefix it with https://bin.bnbstatic.com.
  • captcha# belongs to the token value. Cutting it off makes validation fail.
  • Passing the CAPTCHA doesn't guarantee a login: device and account risk checks still follow.
  • After a few failures the challenge turns into BOX. A fresh session often works better than retrying.
  • The flow is unofficial and changes without warning, so expect endpoint and field names to move.
  • Binance rate-limits and locks accounts quickly. Keep attempts infrequent.

Skip the solver: let BrowserSolver handle Binance

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.