WebSpiderUtils/verification_code

Free

验证码研究破解心得记录。包含网易易盾，阿里云验证码，极验验证码，通用汉字识别，梦幻西游验证等主流验证码破解。包含点按验证码、点选验证、语序点选等等。已更新极验验证码、企业公示网/工商/文书采集系统、极验打码接口。

FreeFree tier

Type

Open Source

Visit Website GitHub (38 stars)

About WebSpiderUtils/verification_code

This open-source GitHub project focuses on cracking various mainstream CAPTCHA systems used primarily in Chinese web services. It covers solutions for NetEase Shield (网易易盾), Alibaba Cloud CAPTCHA (阿里云验证码), GeeTest (极验验证码), general Chinese character recognition, and the popular game Fantasy Westward Journey (梦幻西游) verification. The project includes techniques for click-based, selection-based, and word-order CAPTCHAs. It provides detailed guides on GeeTest JS cracking, point-and-click CAPTCHA character recognition model training, nine-grid classification, and slider trajectory cracking. The repository also contains code demos, training manuals for object detection and recognition models, and a now-closed test API for GeeTest. The project is licensed under MIT and encourages dataset sharing and community contributions.

Key Features

Cracks NetEase Shield (网易易盾), Alibaba Cloud CAPTCHA, GeeTest (极验), general Chinese character recognition, and Fantasy Westward Journey verification

Supports click-based, selection-based, and word-order CAPTCHAs

GeeTest JS cracking workflow to obtain gt, challenge, and validate parameters

Point-and-click Chinese character recognition model training (object detection + classification)

Nine-grid (九宫格) CAPTCHA classification model training manual

Slide trajectory cracking for GeeTest 3.0 and machine learning

Ruishu (瑞数) JavaScript reverse engineering

Includes training manuals for model development with detailed environment setup (NVIDIA-SMI, CUDA, cuDNN, OpenCV, Anaconda) and optimization tips

Previously provided a test API for GeeTest (now closed)

Open-source under MIT license with 1.2k stars and 235 forks

Pros & Cons

Pros

Comprehensive coverage of popular Chinese CAPTCHA types
Provides step-by-step training manuals for object detection and classification models
Includes environment setup guides for machine learning frameworks (CUDA, cuDNN, OpenCV, Anaconda)
Open-source with permissive MIT license and active community (1.2k stars, 235 forks)
Detailed explanations of JavaScript reverse engineering for GeeTest and Ruishu
Encourages dataset sharing and collaborative model improvement

Cons

Test API for GeeTest was hosted on a slow server and is now closed
Some content and datasets are still pending organization and upload
Relies on user-supplied datasets; model training requires significant data collection
Documentation primarily in Chinese, limiting accessibility for non-Chinese speakers
May not keep up with frequent CAPTCHA algorithm updates from providers

Best For

Research and study of CAPTCHA security vulnerabilities in Chinese web applicationsAcademic projects on CAPTCHA recognition using machine learningData collection and labeling for training custom CAPTCHA recognition modelsTesting and evaluating the effectiveness of CAPTCHA systemsReverse engineering and understanding CAPTCHA implementation mechanisms

FAQ

What types of CAPTCHAs does this project cover?

It covers NetEase Shield, Alibaba Cloud CAPTCHA, GeeTest (including point-and-click, sliding, and word-order), general Chinese character recognition, and Fantasy Westward Journey verification.

Is there a working API for GeeTest cracking?

The project previously provided a test API for GeeTest, but it was hosted on a slow server and is now closed. The repository contains code demos for local implementation.

Does the project include model training guides?

Yes, it includes training manuals for Chinese character localization and recognition models (object detection + classification), as well as nine-grid CAPTCHA classification, with environment setup instructions.