Web Authentication Code Generation And Recognition

With further development of internet technology, network is capable to provide more and more information to people. On the other hand, network security issue has drawn more and more attention as well. In order to ensure network security against adverse attacks, most websites have spent great amount of resource to investigate generation technology of verification code.Investigation on generation and recognition of verification code could verify effectiveness of verification code technology to maintain network security. Besides, it will provide convenient service to users to automatically acquire website information, with assistance of Web Agent Program.Topics below will be covered in this thesis:(1) Introduction to principles of verification code technology and one method designed to generate component-free color BMP verification code based on ASP.(2) Description on preprocessing of verification code images including graying, binarization, and denoising of color verification code images.(3) Introduction to segmentation algorithms of verification code images and different segmentation method on two types of verification code images respectively. Improved method of projection segmentation was adopted for verification code in which the spacing of characters is relatively large while fracture exists in the strokes. On the other hand, combination of methods including connected domain, projection, and shortest path was utilized for verification code in which characters are slanted and merged.(4) Recognition of verification code in which characters are greatly distorted using 3 methods including K-Nearest Neighbor Classification, BP Neural Network, and Support Vector Machine in order to analyze effectiveness of verification code regarding network security through recognition rate.(5) Improvement on verification code technology should be made for most websites in terms of network security. The best way for counter-recognition is to increase degree of adhesion between characters so as to make segmentation uneasy.