MSc Thesis Proposal Announcement by Anjali Shah:"Interpreting Biometric-Based Authentication System Reliability Using Biometric Zoos"

Friday, March 6, 2020 - 11:30 to 13:00

MSc Thesis Proposal by:   Anjali Shah

 
Date:  Friday March 6th, 2020
Time:  11:30 AM to 1:00 PM
Location: Lambton Tower, 3105
 

Abstract:  

 
Nowadays, the use of biometrics in authentication systems has increased. Different techniques have been formalized to evaluate such system’s behaviors. One such approach adopted and experimented by many researchers is biometric zoo terminology. We say that the biometric authentication systems based on common machine learning approaches are not that reliable, by experimenting on a keystroke dynamics dataset using the biometric menagerie concepts. In our work we show that the system by itself generates “False Accepts” which is not desirable. Our analysis also says that the users which are “Falsely Rejected” shows some patterns which can be learned to counter such False Rejects. Lastly, we deduced that common machine learning metrics are not useful in finding all the categories of biometric zoo. So, in our further experiments we would like to use anomaly detection with confidence metrics to get all categories if they exist. In keystroke systems, the user gets used to the typing device in short time, so the typing patterns are likely to change over time. To adapt the changing typing behavior a system must have some policy to handle the situation. But How fast the typing pattern changes? In our research, we propose to utilize the formalized biometric zoo categories to learn the keystroke biometric system’s behavior over time.

 

Thesis Committee:

Internal Reader: Dr. Saeed Samet
External Reader: Dr. Ahmed Azab
Advisor: Dr. Sherif Saad Ahmed

 

MSc Thesis Proposal Announcement               

 

5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 csgradinfo@uwindsor.ca