A Research On CAPTCHA Recognition Based On Self-organizing Map

CAPTCHA is now almost a standard security technology. The most widely deployedCAPTCHAs are text-based schemes, which typically require users to solve a text recognitiontask. The state of the art of CAPTCHA design suggests that such text-based schemes shouldrely on segmentation resistance to provide security guarantee, as individual characterrecognition after segmentation can be solved with a high success rate by standard methodssuch as neural networks.In this issue, firstly, we use self-organizing map (SOM) neural network to recognize anon-adhesive CAPTCHA, it is able to achieve satisfactory learning outcomes with lesstraining samples and time, the recognition rate of the CAPTCHA by SOM is almost100%.Compared with other methods,SOM could achieve high success rate with less trainingsamples and less time; Furthermore, we present new character segmentation techniques toattack a text CAPTCHAs of Web QQ, in which characters touch or overlap with each other;we present a novel method to remove the arc or weaken the arc’s effect, which is based onimage skeleton extraction algorithm, and two algorithm to character image segmentation;then input the individual character to the self-organizing map neural network for training orrecognition, the recognition rate is approximately80%.Secondly, we give the CAPTCHA recognition system framework and two sub-moduledesign, introduce the related third-party libraries. Then we present the correspondingexperimental results of these two kinds of CAPTCHAs validation experiments and theiranalysis. The future work and the development trend of CAPTCHA is at the end of theissue.