Enhancing User-Centric Security Using Machine Learning Techniques for Continuous Authentication using Keystroke Dynamic

Abstract

In today’s world, User authentication has gained enormous popularity as of dependency on technological devices. Dependency on the integrity of the user proved to be a tough ask nowadays. Once the user gets access to the system and is authenticated by the system, there are no such criteria to prove whether the user is legitimate or illegitimate. However behavioral biometrics in the context of continuous authentication has achieved remarkable success in this aspect. Advancement in the field of cybersecurity has made it possible to explore new and better ways to defend the user against any cyber attack, including a technique known as continuous user authentication. This approach includes various factors to authenticate users such as Keyboard typing patterns, movement of the mouse, adaptive authentication techniques, etc. However, our primary focus will be to authenticate and identify the user based on keystroke dynamics to continuously relates the user typing pattern(Keystroke) as recorded earlier. Since to capture keystroke dynamics there is no such need for any hardware hence no such hardware is used and software-based technology remained the primary source to gather the data. Like in case, a legitimate financial institution employee logs into the organization’s web portal to access critical financial data and conduct transactions. The system validates her credentials and creates a secure session. The user leaves the account and logs in and an illegitimate user grabs the session by fortune or by applying a phishing/brute force attack and continues with the transactions, in this case, will there be any authentication technique that differentiates both the legitimate and the illegitimate user? The answer lies within continuous authentication by observing the keystroke typing pattern and after sensing variation in the typing rhythm, make the illegitimate user logs out from the session instantly and generates an alarm.