Image verification code generation method and system based on general disturbance resistance

Abstract

The invention relates to an image verification code generation method and system based on general disturbance resistance. The method comprises the steps of obtaining an image set of verification codes, wherein the image set comprises different types of images; generating corresponding general countermeasure disturbance according to all the images in the image set; randomly selecting an image typeas an image to be verified according to the image set; determining questioning information according to the to-be-verified image; screening out a candidate image set from the image set according to the questioning information; adding universal adversarial disturbance to each image in the candidate image set; and determining a verification code test case according to the candidate image set added with the universal adversarial disturbance and the questioning information. According to the invention, the automatic identification attack of the convolutional neural network can be resisted, and thesecurity of the image verification code is improved.

Description

technical field [0001] The invention relates to the field of image verification codes, in particular to a method and system for generating image verification codes based on general anti-disturbance. Background technique [0002] Captcha, also known as the Turing Test (Completely Automated Public Turing Test to Tell Computers and Humans Apart, CAPTCHA), is a computing program that distinguishes whether a user is a computer or a human. The program generates a test that humans pass easily but computers struggle to pass. The purpose of the verification code is to prevent the automatic attack of the computer system. For example, users often use the verification code in the user login or message system. After entering the user name and password or entering the message, they need to further enter the verification code, which can effectively avoid the use of Malicious behaviors such as automatic computer brute force password cracking, automatic ticket swiping, automatic posting, et...

AI Technical Summary

Problems solved by technology

However, with the development and application of deep learning, the ability of the convolutional neural network model to recognize images has been greatly improved and even surpassed that of humans, which makes it possible to automatically identify attacks based on deep learning computer programs and can attack the current verification code protection. Therefore, the traditional image verification code is no longer safe and cannot prevent automatic attacks through computer programs, such as brute force password cracking, automatic ticket swiping, and automatic posting, which pose a serious threat to the security of the website and users.

Images