Using image comparison algorithms to detect malicious zero-day apps and domains

Malicious actors frequently use fake sites or applications to collect sensitive data from users. In our research, we focused on situations when an attacker uses an icon of a trusted service or application to trick a user into taking an action he would not take. It could be a malicious version of a popular application, or a phishing site that mimics a login page to a well-known service. In this session, we will look at how to use icon similarities using image comparison algorithms of neural networks to improve zero-day application and site detection.