Robust Adversarial Example as Captcha Generator
Nicolaus Boby Ardhita, Nur Ulfa Maulidevi · 2020
Adversarial example as one of the weaknesses of Deep Learning, can be used to attack image recognition models. Image recognition model can be used to attack common image CAPTCHA schemes, as the AI algorithm can recognize images to pass the test. So with applying robust adversarial example to image CAPTCHA, we produced a way to synthesize a robust CAPTCHA generator that are robust and cannot be easily attacked by image recognition algorithms.