Video Data Collection for Continuous Identity Assurance
Frequently monitoring the identity of a person connected to a secure system is an important component in a cyber-security system. Identity Assurance (IA) mechanisms which continuously confirm and verify users’ identity after the initial authentication process ensure integrity and security. Such syst...
| Main Author: | |
|---|---|
| Format: | Others |
| Published: |
Scholar Commons
2016
|
| Subjects: | |
| Online Access: | http://scholarcommons.usf.edu/etd/6424 http://scholarcommons.usf.edu/cgi/viewcontent.cgi?article=7620&context=etd |
| Summary: | Frequently monitoring the identity of a person connected to a secure system is an important component in a cyber-security system. Identity Assurance (IA) mechanisms which continuously confirm and verify users’ identity after the initial authentication process ensure integrity and security. Such systems prevent unauthorized access and eliminate the need of an authorized user to present credentials repeatedly for verification. Very few cyber-security systems deploy such IA modules. These IA modules are typically based on computer vision and machine learning algorithms. These algorithms work effectively when trained with representative datasets. This thesis describes our effort at collecting a small dataset of multi-view videos of typical work session of several subjects to serve as a resource for other researchers of IA algorithms to evaluate and compare the performance of their algorithms with those of others. We also present a Proof of Concept (POC) face matching algorithm and experimental results with this POC implementation for a subset of collected dataset. |
|---|