Skip to main content

CAPTCHA guide

Akamai Bot Manager

Edge-level bot detection that scores sessions with a JavaScript sensor before anything is shown.

Difficulty: Very high
Risk engine
Exploded concept drawing of how the Akamai challenge is built

Akamai Bot Manager is an enterprise bot-detection layer in front of many large retail, banking and travel sites. Most visitors never see a CAPTCHA. Akamai scores the session and then serves the page, runs a JavaScript sensor challenge, or returns an "Access Denied" page. When a visible CAPTCHA does appear, it is usually a separate product added on top.

How to identifyLink to How to identify

  • Cookies named _abck, bm_sz, ak_bmsc, bm_sv, bm_s, bm_so and sec_cpt
  • An _abck value split by ~, showing ...~-1~... before validation and ...~0~... after
  • A sensor script on a first-party path or under /_bm/_orchestration_/
  • A POST whose JSON body looks like {"sensor_data":"2;...;<ciphertext>"}
  • Server: AkamaiGHost on blocked responses
  • An "Access Denied" page that shows a reference ID
  • An Akamai-BM-Telemetry header on some API endpoints

How it worksLink to How it works

  1. Akamai rates the first request on its IP, TLS fingerprint, headers and cookies.
  2. The page pulls in a sensor script that collects browser and device signals.
  3. The script posts sensor_data back to the edge, and the reply updates _abck.
  4. A validated _abck carries ~0~ in its state field. ~-1~ means it has not been validated yet.
  5. Protected endpoints expect a valid _abck and may also check bm_sz, bm_sv or a telemetry header.
  6. Some sites add SBSD / Content Protector, which slows suspicious clients down (tarpitting) and sets extra cookies (sec_cpt, sbsd).
  7. A 201 {"success":true} after a sensor post does not prove anything. Always check _abck and a protected request.

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

Tool What it does Success Rate Link
Patchright / Camoufox Stealth browsers to run the sensor and warm up cookies 40–70% Kaliiiiiiiiii-Vinyzu/patchright-python · daijro/camoufox
DrissionPage A real Chromium session where cookies are easy to inspect 35–65% g1879/DrissionPage
SeleniumBase UC mode Selenium with patches, for pages running the JS sensor 35–60% seleniumbase/SeleniumBase
curl_cffi Imitates browser TLS/JA3 once cookies are warm 25–50% lexiforest/curl_cffi
nodriver CDP automation that leaks less by default 30–55% ultrafunkamsterdam/nodriver

Practical approachLink to Practical approach

  1. Warm the session in a real browser: open the homepage, let the sensor finish, and keep every cookie.
  2. Look at _abck after the sensor post. Wait for the ~0~ state, or test with a protected request.
  3. Send API calls with the same cookies, User-Agent, Accept-Language and client hints.
  4. Space requests out like a person would. Sensor posts fired back to back are a clear bot signal.
  5. On sites using SBSD / Content Protector, expect deliberate slowdowns and avoid aggressive retries.
  6. For login and checkout, a full browser session beats raw HTTP.

Practical tipsLink to Practical tips

  • Treat _abck as opaque: send it back exactly as you received it and never rebuild it.
  • The sensor relies on bm_sz. Losing it can break the session.
  • Cookie names and formats differ between customers, so save the whole jar by name.
  • A validated cookie can lose its standing later, so keep the session's behaviour consistent.
  • Homepages are often lightly guarded. Test against the endpoint you actually need.
  • Akamai sometimes runs next to another vendor (watch for __uzm* cookies). Don't assume there is only one system.

Skip the solver: let BrowserSolver handle Akamai

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.

Akamai unlock guide →

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.