Using Generative Adversarial Networks to Break and Protect Text Captchas
- Submitting institution
-
The University of Leeds
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- UOA11-4486
- Type
- D - Journal article
- DOI
-
10.1145/3378446
- Title of journal
- ACM Transactions on Privacy and Security
- Article number
- 7
- First page
- -
- Volume
- 23
- Issue
- 2
- ISSN
- 2471-2566
- Open access status
- Compliant
- Month of publication
- April
- Year of publication
- 2020
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
8
- Research group(s)
-
E - DSS (Distributed Systems and Services)
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Text captcha schemes are one of the most widely used web security systems. This work presents new algorithms, based on generative-adversarial deep learning, for automatically solving text captchas. The proposed scheme was the most effective text-captcha solver in terms of accuracy and training cost, defeating all text-captcha schemes used by the top-50 of the world’s most popular websites. It also proposes countermeasures for the attack. Results were first published in ACM CCS 2018 and nominated for the Best Paper, as one of the top-4 ranked papers selected from 804 submissions(top 0.5%). Received coverage on 100+ media outlets, including ZDNet.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
