The relevance of the problem considered in the article is due to the growing number of web resources of various government organizations and commercial companies on the Internet, and one of the reasons for data leakage may be mass collection of information. The information obtained during the collection can become a tool in the capable hands of an attacker. Attackers can easily create and upload malicious software with information collection functions to the network, and the collected information can be used to carry out attacks with the power of social engineering methods. The most suitable methods for such attacks are phishing and pretexting. The article provides an overview of the problem of illegal mass collection of information and its use by hackers. Possible methods of countering mass data collection are analyzed, options for creating captchas are considered, and their disadvantages are the possibility of recognition and guessing by an attacker or bot. As a result, an improved method is proposed that solves this problem. This paper describes the main functions of the program, as well as possible variations in the use of captcha generation. To protect against captcha recognition by a trained bot, it is suggested to enter a semantic load into the image. As a result, we developed and presented a method for creating a captcha that is resistant to automatic text recognition.