Activity Based Verification - Phase 2
The project Activity Based Verification aims to recognize a user based on its specific behavior. Thereby the person to be recognized will not be incriminated and the use of expensive hardware will be avoided. In addition, the user’s identity will be verified at any time.
It is clear that a person can be represented by his biometric features. In daily life, hand written signatures have been widely accepted. Despite of the fact that a majority of the human can not differentiate a falsified signature from a genuine one, it carries very clear nuances (e.g., pressure, speed, etc.) which can not be imitated and using which a trained specialist can differentiate them clearly. “Activity Based Verification” follows this direction. Similarly, a user’s behavior in using computer, peripherals and certain applications can not be imitated by another person; therefore, can be used to identify the user. This can be regarded as a behavior-based biometric characteristic. The biometric characteristic is always carried with the user and cannot get lost or stolen. The conscious or unconscious reveal of such an authorization is clearly not possible. The challenge here is to capture the behavior of a user and to map it in machine evaluable patterns. This in turn requires selecting proper user generated events using which it is possible to differentiate users based on his behaviors. It is possible to consider three types of user generated events for detecting user behavior. First of all, the mouse events include a wide range of essential information (e.g.: clicks, movements, speed, etc.) which can be used generate a virtual signature for the user.
Second, in a similar fashion, keyboard events can be used to verify a user (e.g. input speed, frequency, combinations of selected keys). Third, application use provides useful information about the preferences and thus the behavior of a user. In this project, we are both extending the existing research activities and developing combined approaches for improved results. Our objective is to capture different level of events to derive a context and understand the characteristics of a user. Similar to a hand made signature signed by a person as it is unconsciously realized (if we mask the will to sign something); framework proposed here will continuously and transparently verify user authenticity based on his unconscious activities. It will enable us to continuously ensure that the user is still the authenticated one. Simultaneously, the solution will be designed to be resistant against attacks such as replay attacks, Trojan horses, key loggers, social engineering methods, etc. In addition to a prototypical implementation, empirical analyses of the usability of the framework for different scenarios will be provided. Guidelines will be elaborated for service providers, which could offer these verification methods to their customers.