USING ADVERSARIAL ATTACKS TO PROTECT AGAINST AUTOMATIC TEXT RECOGNITION IN CAPTCHA SERVICES
I. S. Tsurkan, E. V. Malyshev, E. A. Zubkov, D. A. Moskvin Peter the Great St. Petersburg Polytechnic University
Annotation: The paper proposes a method of CAPTCHA generation using adversarial attacks with respect to the image recognition model. The advantages of this method are determined based on comparative analysis of statistical data obtained as a result of the study. The developed method allows increasing the efficiency of counteraction against automatic bypassing of CAPTCHA services due to the use of insecurity of neural network models from malicious influence
Keywords: adversarial attacks, text recognition, denial of service, CAPTCHA services, machine learning, Python
Pages 165-175