This repository provides a collection of projects and resources focused on the analysis and cracking of various verification code systems. The project addresses the challenge of automating the recognition and bypassing of these security measures, particularly those used in online services and games. The core techniques involve image processing, machine learning (especially deep learning), and web scraping to extract necessary parameters. The primary technology used is Python with deep learning frameworks like darknet.
This project offers a comprehensive collection of tools and techniques for tackling different types of verification codes, from image-based to interactive. It emphasizes practical implementation and includes detailed documentation on data preparation, model training, and inference. The repository is continually updated with new methods and solutions, offering a valuable resource for security researchers and automation enthusiasts. The inclusion of datasets and training guides makes it a self-contained learning and experimentation environment.
- Jiyuan Verification Code Cracking: Provides JavaScript decryption logic for obtaining validation parameters from the Jiyuan verification system. - Character Recognition Models: Includes models for recognizing Chinese characters from verification codes, specifically addressing dot-click and selection types. - Nine-Grid Verification Codes: Covers data preparation and training guides for classification models to handle nine-grid verification challenges. - Machine Learning for Sliding Verification: Focuses on analyzing and potentially automating the process of interacting with sliding verification systems, including tracking user movement. - Data Collection & Preprocessing: Offers resources for acquiring and preparing datasets for training verification code recognition models. - Model Training and Evaluation: Provides comprehensive training manuals with datasets and guidance on evaluating model performance. - Python Integration: Presents examples and documentation for using trained models within Python applications.
The project is in an active development phase with ongoing updates and additions. The core components, especially for Jiyuan and character recognition, are relatively mature and well-documented. However, some sections, such as sliding verification, are still under development. While the project has a dedicated community presence and regular commits, the reliability of certain services (e.g., temporary testing interfaces) should be considered. The project benefits from comprehensive documentation, including installation guides, training manuals, and code examples.
This project is valuable for security researchers, automation developers, and anyone interested in understanding and overcoming verification code challenges. It provides practical tools, datasets, and documentation for automating verification processes, which can be beneficial for testing, data scraping, or bypassing security measures in controlled environments. The resources offered can significantly reduce the effort required to handle various verification code types compared to manual methods.