reports that the novel machine learning-based CAPTCHA solver dubbed DW-GAN yielded a 94% success rate in addressing dark web challenges, which was higher than image-level convolutional neural network alone, image-level CNN with processing, and character-level CNN with segmentation.
DW-GAN, developed by researchers at the Universities of Arizona, South Florida, and Georgia, leverages background 'denoising' and character segmentation and recognition to differentiate letters and numbers in CAPTCHA images.
Researchers tested DW-GAN on the defunct dark web market Yellow Brick and found that the solver was able to gather market intelligence from the site in nearly five hours and solved all CAPTCHA challenges within 18.6 seconds.
"Overall, the proposed framework could automatically break CAPTCHA with no more than three attempts. Breaking all CAPTCHA images take about 76 minuets [sic] in total for all 1,831 product pages, a process that is fully automated," said researchers.
The solver, which has already been uploaded on GitHub
, may help in combating cybercrime, added researchers.