基于深度学习的人工智能用于识别破解字符型验证码  

Crack CAPTCHA by AI based on Deep Learning

作　　者：周正 文亚飞 鲍文平 

机构地区：[1]腾讯公司,广东深圳518000

出　　处：《通信技术》2017年第11期2572-2576,共5页Communications Technology

摘　　要：作为互联网行业通用的基础安全策略,验证码担负着保护网络帐号安全体系的第一道防御,用以验证和防范机器程序的恶意行为。近年来,不法分子利用基于"深度学习"的人工智能技术搭建打码平台,快速海量识别破解字符型验证码,为网络黑产分子撞库窃取用户信息作恶提供帮助。因此,以腾讯守护者计划安全团队协助警方破获首例AI人工智能犯罪案件为例,对基于深度学习的人工智能技术用于识别破解字符型验证码问题进行解析,以期为提升和优化互联网行业安全策略提供借鉴,为互联网生态安全建设提供参考。As the basic security strategy of the Internet industry, CAPTCHA is the first line of defense responsible for protecting the security of the account system in Internet, and preventing the attack from malicious behavior of the automatic machine program. In recent years, there arises some CAPTCHA-CRACK platforms based on the latest artificial intelligence technology of deep learning, it could crack the CAPTCHA systems with high efficiency and result in a large number of user private data stolen. For this reason, and with the case that Tencent Guardian program security team assists the police to uncover the first AI artificial intelligence crime as an example, the deep learning-based artificial intelligence technology is used to identify CAPTCHA-CRACK, expecting to improve and optimize the Internet industry security strategy and provide reference for the construction of Internet ecological security.

关 键 词：验证码 卷积神经网络 人工智能 深度学习 多标签 

分 类 号：TN918[电子电信—通信与信息系统] TP181[电子电信—信息与通信工程]