Research On Security Assessment Of CAPTCHA Recognition Based On Cloud Platform

The CAPTCHA is an effective way for authenticating the user to be a person rather than bot programs,which can effectively guarantee the security of the network.With the development of deep learning and machine vision,many of the text-based CAPTCHAs with the security protection technology have been broken.Rotation adhesion hollow CAPTCHAs with three-security protection technology,as a novel CAPTCHAs,can be said to be the optimization of solid CAPTCHAs.The connected hollow characters using antisegmentation technique are formed by the contour lines,which are considered as high security by many Internet sites.Whether the security protection technology can effectively improve CAPTCHA security and enable users to have a better user experience is not yet perfect.Therefore,our research proposes a CAPTCHA recognition based on cloud security evaluation,which will capture many captcha images stored in the cloud platform and use three evaluation models to evaluate the CAPTCHA recognition.Firstly,the present paper carries out the research of CAPTCHA security protection technology and propose three evaluation models.Secondly,three assessment models are used to evaluate the rotation adhesion hollow CAPTCHAs with three-security protection technology.The evaluation process includes preprocessing,single character recognition and whole captchas identification.Finally,this paper sets up a CAPTCHA recognition system based on the cloud platform,which will capture the CAPTCHA from the the Internet to store on the cloud platform and use the flask framework to build a CAPTCHA recognition system through users' feedback to identify the error label,changing and increasing the training set to improve the CAPTCHA recognition ability.This paper uses three evaluation models successfully breaked the adhesion of the hollow CAPTCHA.The end-to-end security assessment model has good recognition effect on the distorted adhesion of the hollow CAPTCHA recognition and does not require the complicated pretreatment and character segmentation work,which is universally applicable for text-based CAPTCHA recognition.