atbCAPTCHA is a behavioural risk-control product. Depending on its risk score it shows a click, slider or image-alignment task, or nothing at all. It is found in finance, telecoms, online retail and IoT, and it checks traffic in several stages instead of relying on one puzzle.
Challenge typesLink to Challenge types
- Slider: line up a piece or a cut-out
- Image click: click or align the requested images
- Invisible: a token is issued silently to low-risk sessions
- Multi-stage validation: more checks run after the visible task
How to identifyLink to How to identify
- Global
appIdandapiServervalues in the page source apiServerfrequently set tohttps://cap.aisecurius.com- SDK setup with a
successcallback:as.Captcha(element, {appId, apiServer, success}) - A token handed to that callback, often a short prefix followed by a colon
- atbCAPTCHA branding, or a widget served from the
apiServerdomain - A hidden form field or JS callback that receives the final token
How it worksLink to How it works
- The page starts the widget with
appIdandapiServer. - The client asks for a challenge, and atbCAPTCHA picks its type from the risk score.
- The visible task, or the invisible check, produces a token.
- The token reaches the site's
successcallback and is sent to the backend. - atbCAPTCHA validates it along with the IP, User-Agent, timing and behaviour.
Free bypass methodsLink to Free bypass methods
| Method | What it does | Success Rate | Link |
|---|---|---|---|
| DrissionPage | Runs the widget in real Chromium and catches the callback token | 70-85% | g1879/DrissionPage |
| OpenCV | Finds the slider gap via template matching or edges | 70-90% (visible slider) | opencv/opencv-python |
| ddddocr | OCR and object detection for image tasks | 65-85% | sml2h3/ddddocr |
| YOLOv8 | Detects target images in click and align tasks | 70-90% (with a trained model) | ultralytics/ultralytics |
| Patchright / Camoufox | Stealth browsers that keep the risk score down | 60-85% | Kaliiiiiiiiii-Vinyzu/patchright-python |
| undetected-chromedriver | Selenium with detection patches | 60-80% | ultrafunkamsterdam/undetected-chromedriver |
Practical approachLink to Practical approach
- Note the
appId, theapiServerand the page URL. - Load the page in a real browser on a clean residential IP.
- When the check is invisible, simply wait for the
successcallback. - When a slider shows up, solve it with OpenCV and drag the way a person would.
- When an image task appears, find the requested pictures and click them in order.
- Take the token from the callback or the hidden field and submit right away.
Practical tipsLink to Practical tips
- The token is linked to the IP and User-Agent that earned it. Solve and submit through the same proxy.
- Validation has several stages: a correct puzzle answer still fails if the fingerprint looks wrong.
- Datacenter IPs are usually rated high-risk. Use residential addresses.
- Reloading the widget after solving can void the token.
- Risk-adaptive systems watch how you interact, not just the final answer, so keep mouse paths natural.