Research on Image Verification Code based on Watermarking Technology

With the development of the Internet, security on networked computers is becoming more and more necessary. Identity verification codes provide additional security for web systems when visitors log in. Among various verification code technologies, dynamic picture verification code has become the most widely used. With the development of machine learning and neural network algorithm, automatically identify information from verification code become convenient for machines. Meanwhile, multimedia security has become an important field of research about watermarking, which is a technique that embeds some identity information into audio or image, and then the embedded information can be restored for different purposes. In this work, an imaging technology combining DWT(Discrete Wavelet Transform), SVD(Singular Value Decomposition), and DCT(Discrete Cosine Transform) is applied to the field of image verification code. We embed image caption as a watermark into its corresponding image. When authenticating users, the watermark will be extracted and the mapping relationship between watermark and image will be shuffled based on rules. When the mouse is over one image, the shuffled caption is displayed as clue words. Only when the visitors select all image with correct mapping relationship, it will be validated. This method can effectively distinguishs whether the visitor is a person and prevent computers identifying.