Researchers at Lancaster University in the U.K., Northwest University, and Peking University in China have demonstrated a deep learning algorithm that could render captcha security and authentication redundant.
The algorithm solves captchas with substantially greater accuracy than earlier captcha attack systems, and successfully cracks captcha versions that defeated previous hacks.
The system uses a generative adversarial network (GAN), educating a captcha generator to produce large numbers of training captchas that are indistinguishable from actual captchas.
These are employed to quickly train a solver, which is tested against real captchas; the algorithm only needs 500 genuine captchas, rather than the millions required to train a conventional attack program.
Lancaster's Zheng Wang said, "Our work shows that the security features employed by the current text-based captcha schemes are particularly vulnerable under deep learning methods."
From Lancaster University
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA