A method and system for generating character adversarial verification codes based on frequency-domain noise addition

Abstract

The invention discloses a method for generating a character confrontation verification code based on frequency-domain noise, comprising the following steps: generating a string of character sequences by a random algorithm, and sequentially converting each character in the character sequence into a character sequence containing a single corresponding character. Character image; convert the character image from the spatial domain to the frequency domain, add anti-noise to the character frequency domain image according to the method of adversarial learning, and then convert the character frequency domain image after adding the anti-noise from the frequency domain to the spatial domain; Each character image after adding anti-noise is spliced ​​together in the order of the verification code answers to form a character anti-noise verification code. The invention also discloses a generation system of character confrontation verification code based on frequency domain noise addition. Based on the confrontation learning method, the invention adds confrontation noise to the character image in the frequency domain, and the generated character confrontation verification code can effectively avoid or weaken the image preprocessing adopted by the attacker, and improve the security of the character verification code.

Description

technical field [0001] The invention relates to the fields of machine learning model security and adversarial verification code security, in particular to a method and system for generating a character adversarial verification code based on frequency domain noise addition. Background technique [0002] The verification code technology is mainly to be able to distinguish whether the current user is a normal human being or a machine. As a relatively common form, character verification codes have good application value in practical applications such as website login and email verification. [0003] The usual character verification code design needs to follow two principles: 1) Human identifiability: Since the main function of the verification code is to distinguish whether the user is a human or a machine, human users must be able to easily identify the content in the verification code. This requires that the characters in the character verification code cannot be disturbed to...

AI Technical Summary

Problems solved by technology

This requires that the characters in the character verification code cannot be disturbed too much, otherwise it will greatly reduce the experience of real users

2) Machine unrecognizability: the characters in the character verification code need to be difficult to be cracked by automatic programs

Therefore, it is unwise to directly abandon the research of character verification codes

Images