Machine Learning Methods in Authentication Problems Using Password Keystroke Dynamics

Abstract

We examine the problem of user authentication from keystroke dynamics. A new static authentication method that collects information about user keystrokes is described. Its applicability to the authentication problem is substantiated by experiments and the optimal conditions for the implementation of the method are chosen.In most modern information systems, one of the main issues, alongside with data storage and processing, is the granting of access permissions. Permissions prevent unauthorized access by outsiders and also differentiate between the access rights of staff members within the organization. The authentication problem, i.e., verifying the legitimacy of the user seeking to access the system, is thus one of the key problems in information systems.Various methods are available to solve this problem. The most popular technique used in modern information system is password authentication -authentication by a special string of symbols known only to the user who has permission to access the system. In addition to obvious advantages -passwords are simple to implement and their use is very widespread, password authentication suffers from essential shortcomings: the user may forget the code string or reveal it to other users; if the password is too short or too simple, it can be found by direct enumeration or by dictionary-based enumeration. All this casts serious doubts on password use in high-security systems.Authentication by means of electronic and digital signatures (EDS) is free from the last shortcoming [1]. However, it is impossible to ensure secure storage of closed keys in this way, because they are stored using other access control tools.For these reasons, many effective security systems rely on biometrics to identify the user. Biometric systems fall in two categories. One category includes systems using features that differ for natural reasons among different individuals: fingerprints, retina scans, voice patterns, body temperature chart, etc. These systems are effective but quite costly because they require special equipment. The second category includes systems that analyze user behavior; they are based on user experience and skills. They do not require any special equipment and are simple to introduce. One of such systems analyzes user keystroke dynamics.