That’s not you! Applying Neural Networks to Risk-Based Authentication to Detect Suspicious Logins

Authentication mechanisms are an essential component of digital security, safeguarding sensitive data from unauthorized access, preventing unauthorized purchases, and hindering impersonation. Among these systems, risk-based authentication (RBA) is an advanced method that enhances security by evaluating contextual information in addition to user credentials. Our research explores the use of machine learning models to improve the classification performance of RBA systems. Using a set of publicly available login data, consisting primarily of IP addresses and user agent strings, we train and evaluate a feed-forward neural network (FNN). Compared to previous research, we observe substantially improved accuracy for login classification. Our findings demonstrate that FNN-based RBA effectively decreases re-authentication rates for legitimate users, thus making it an interesting and promising area for further RBA research.