Skip to main content

Multi-Modal Biometric Sensor Networks

Researcher: Marios Savvides

Abstract

Multi-Modal Biometric Sensor Networks for User Identification/Authentication

This proposal is aimed to increase the research thrust of Multi-Modal Biometric recognition research for secure Biometric authentication to physical infrastructures on CMU Campus (more specifically on the CyLab building). We propose to install several Biometric Sensor Network nodes which will compromise of a PTZ (pan-tilt-zoom camera, fingerprint sensor (with added ability to read smart-cards) and an Iris acquisition camera. Possible locations include at (or near) CyLab building entrance and at other indoor access-points; we plan to install (at least one Multi-Modal Sensor node) outside an office for allowing access inside that office. Currently most authentication systems rely on single Biometric modalities to perform identification and verification, however it is possible for an attacker to spoof a single biometric to gain access but extremely hard for the same attacker to spoof all three different Biometric modalities (face, fingerprint and iris) simultaneously. This is the main motivation behind this project and how to successfully deploy it within CyLab building (theory part + implementation practical part to create high profile demos).

The MultiModal Biometric Sensor nodes will be designed to ease their deployment to different locations (depending on application scenario). The sensors can be designed (including building custom hardware enclosures with enclosed mini-single-board computers) such that they are completely independent and have wireless and Bluetooth access to be able to interface with the Sensor Network developed by Prof. Mike Reiter and Adrian Perrig (for accessing other building sensors and controlling the office electronic door-locks for physical access).

Proposed Research includes evaluate Sensor Deployment at:

1.        CyLab building entrance: Pan-Tilt-Zoom camera (network camera which only requires Ethernet connection and power), it is possible to use 802.11b wireless but frame rate acquisition will be slow ~5 fps. The PTZ camera will have the ability to scan and perform surveillance at the entrance. It will have the ability to detect human people, and detect faces in a scene, zoom in and track human faces of people walking towards the entrance and perform identification and authentication by the time the visitor reaches the door!

This preliminary identification step can also reduce overall verification time as it will retrieve the remaining biometric templates (fingerprint and iris) of the user, thus all that remains when the user reaches the door is a combined Multi-Modal Biometric verification (a one-to-one) matching which is computationally very fast. If the user’s fingerprint, iris and face had to be searched in the database (to see if it matches with any of the enrolled users) that could take significant computation and time before giving access. (This mode of operation is to completely remove the need for authorized CyLab building user’s to carry swipe cards or any other form of IDs that can be stolen or forgotten). However we will have the option and infrastructure to accept smart-cards which will carry the user’s biometric templates. When the user reaches the door, he/she will insert their smart-card into the reader (which will say to the system ‘I am John’); the system will then verify based on the stored biometric templates whether indeed he is John using his face, fingerprint and iris pattern.

2.        Outside a CyLab office: We will also setup an identical Multi-Modal sensor node outside at least one CyLab office (and more when the initial system is successful). Camera will be mounted on top of the office door and a fingerprint sensor on the wall next to the door. We also plan to build a Desktop Multi-Modal authentication system: This will serve as a demonstration of Multi-Modal Biometric Authentication system to virtual spaces (allow screensaver log-in / secure file access / decryption). This project is challenging as it requires good synchronization and control of multiple biometric sensors (face, iris and fingerprint) as well as other secondary devices such as smart-cards (which can be used as a means to store biometric templates). Sensors and Sensor SDKs will need to be purchased to allow efficient programming and acquisition of Biometric data from each Biometric sensor modality.