Text-based CAPTCHA Cracking Based on BP Neural Network
Yuan Gao · Electronic Science and Technology · 2012
This paper introduces the method for cracking the CAPTCHA of one game site in detail.Our work is mainly divided into three parts:character extraction,character correction and character recognition.In the character extraction phase,we need to preprocess to reduce the extraction difficulty and this phase mainly adopts similar color statistics;in the character correction phase,an improvement is made of the traditional rotating to normalize the characters;and in the character recognition phase,BP neural network method is used with a success rate of 70% and an average time of cracking the web site of 1.625 s.