BEHAVIORAL-BASED AUTHENTICATION FOR BUSINESS ACCOUNTS USING HIDDEN MARKOV MODELS ON MOUSE MOVEMENT DATA
Roberto Acevedo Facultad de Ingeniería
Keywords: Behavioral-based authentication, HMM, mouse movement data, user behavior modeling, continuous authentication, etc.
Abstract
One new technique that is coming up is behavioral-based authentication where users can be identified by their specific behavioral patterns which here include the movements of the mouse. This paper concentrates on the vision of using Hidden Markov Models (HMM) to study the behavior of the mouse to know how to get business accounts. Common authentication systems such as passwords and tokens can be stolen, phished, and re-played. Biometric systems, more secure though capable of being spoofed, are not able to authenticate users continuously as they are in active sessions. In order to overcome these constraints, we introduce a new model known as Mouse Behavior Authentication with HMM (MBA HMM) which takes advantage of the statistical strength of the HMM to model and identify the actual user behavior using mouse trajectory data. The model includes pre-processing raw information on mouse movements, deriving both temporal and spatial information, and learning user-specific HMMs to identify abnormalities in real-time. To detect genuine users and non-genuine users in business applications, MBA-HMM is constantly tracking the patterns of mouse movement. It is a passive and unobtrusive agent that works in the background and improves the security without interfering with the user experience. The experimental and testing of the proposed method on actual mouse movement data reveals that it has a promising future. MBA-HMM performed well in user identification and resisted imitation attacks as well as false acceptance and rejection. These results indicate how effectively behavioral biometrics, specifically, mouse dynamics based on HMMs can be used as a strong solution to the need to improve the security of business account authentication systems.