eVALUATING Biometrics

Sep 1, 2004 12:00 PM, BY RAM SATHAPPAN

Larry Barfield recently found himself charged with the task of evaluating fingerprint-based biometric security for the U.S. Marine Corps. As a project engineer for the Space and Naval Warfare Systems Center (SPAWAR), Barfield had a strong desire to increase the reliability of high security sites. However, he didn't know how to integrate fingerprint identification technology with legacy access control systems, nor how well it would work, what it would cost or how it would be accepted by users. To learn the answers Barfield decided to design and install a six-month trial system at a high-traffic site.

COMBINING CARDS AND FINGERSCANS

Barfield opted to evaluate the V-Smart system from Bioscrypt, which allows users to carry their biometric information identities with them. Instead of storing fingerprints in a centralized database, users carry a contactless smart card, which holds a template derived from a fingerscan of one of their fingerprints. The combination of a contactless card and fingerscan provides two-factor authentication methods to provide an added assurance of reliability and security.

The distinction between a fingerscan and fingerprint is an important one. Instead of storing the fingerprint, the Bioscrypt system converts the scanned fingerprint image into a fingerscan template. To match a fingerprint to a fingerscan, the reader compares the live version of the scanned fingerprint to the data within the master template stored on the contactless smart card. Because the fingerscan template was created using pattern-based technology, which has a lower-resolution, smaller-area and reduced information content than a full image, the original fingerprint cannot be re-created with the full fidelity of a typical forensic-quality fingerprint. Furthermore, the pattern-based template is quite different from the minutiae-based templates that are typically used with forensic systems.

Moreover, since the fingerscan is stored on the card, there is no need for a central database of private and sensitive biometric information that can be abused or compromised. The card identifies the individual and his or her access rights, much as the installation's legacy MDI SAFEnet magnetic stripe access control system identified users. With the SAFEnet system, however, anyone who knew the PIN associated with a particular identity card could exercise the permitted access rights. With the Bioscrypt system, these rights are granted only if the card holder can produce the fingerscan associated with the card. It is impossible, therefore, for someone to steal and use the card without the associated user present.

The site ultimately chosen for the trial had a population of 735 people. “We ran into several things we didn't expect,” Barfield says. “This facility has personnel who work on aircraft and have their hands covered with dirt or oil. The only real problem we ran into, however, was on rainy days when people didn't dry their fingers first before scanning them.

“We were also surprised by the change in throughput,” Barfield continues. Magnetic stripe system users experienced an average delay of 1.9 seconds between the time they entered their PIN to when the solenoid on the turnstile released. With the Bioscrypt card/fingerscan approach, throughput dropped to 1.7 seconds, a 10 percent improvement. “You don't usually increase security and improve throughput at the same time,” he said.

THE COMMERCIAL APPROACH

The installation's commercial approach is very different than traditional deployments of biometric technology that take what is known as a “centralized” or “forensic” approach to identifying users. For example, a user's fingerprint is collected and then matched against a database of known fingerprints and associated identities. In the case of a building with multiple entry points, each access point must have real-time access to a centralized server managing this database. Additionally, as the number of users increases and there are more entries to check against, the time-to-match increases, as does the probability of a false-accept, where the system permits access to a person under the wrong identity, since each user must be checked against all other users in the system.

Because the database is shared, these types of systems require significant processing and storage resources. If several access points request verification at the same time, as is typical during heavy-usage periods such as lunch or end of the work day, the average throughput time per user increases. Forensic approaches also carry a negative stigma because user fingerprints are stored in a database; users either feel like criminals when their prints are collected or are left wondering how secure the database — and their privacy — really is.

SHARING THE LOAD

What makes a commercial approach work more efficiently than a traditional, forensic approach is the shared labor between the access point and the central server. In a forensic system, the access point captures the fingerprint, has the potential to do some minor processing, and then send it to a central server. The central server has to match the fingerprint with a high degree of confidence to a single entry in its biometrics database. The server is also tasked with searching a second database to establish access rights. With multiple access points and a large database, an operator needs a very powerful and expensive system to be able to keep the average throughput at a reasonable level.

Compare this system to a commercial approach where the access point compares the live fingerscan with the master fingerscan template on the card instead of having to match a single fingerscan to a database. The access point itself compares the live fingerprint with the fingerscan template stored on the card. By using a more intelligent access point, the server has much less work to do, regardless of how many users there are. For example, Bioscrypt is involved in a project with more than 50,000 users. Throughput for a forensic system with a database this size would be greatly reduced. Using the commercial approach, throughput is about the same, whether there are 150 or 50,000 users.

In Barfield's case, the central server remained unchanged. The primary hardware modification dealt with the card readers themselves, which had to be exchanged to accommodate the new fingerprint scanners. This upgrade potential is key to the very simple integration of the biometric technology.

Also key to the Bioscrypt system are its pattern-matching algorithms and digital signal processing (DSP) technology from Texas Instruments. Bioscrypt selects from a wide range of DSPs for its access points, depending upon the needs of the particular installation. DSPs are able to process images more quickly and efficiently than other kinds of processors and provide extensive image enhancement capabilities. As a consequence, Bioscrypt enhances the fingerprint image captured from the fingerprint scanner, “cleaning” up the fingerprint, so to speak, before converting it to a fingerscan. This cleaner fingerscan improves matching reliability and accuracy, which accounts for an almost non-existent false-reject rate.

WORKING IN THE REAL WORLD

“It was vitally important that we could easily integrate the Bioscrypt system with our existing access system,” Barfield says. “We found that it was mostly a matter of configuring the card data format for each of the systems to work together. As a consequence, we completed the project on time and approximately 20 percent under budget.

“The trial, in my opinion, was a success,” he continues. “We wanted to evaluate biometric technology and discovered that we could adapt this new technology to our existing system to create a more reliable access control infrastructure that not only gives us a higher security posture, but also reduces throughput time.”

A critical part of the trial's success stemmed from the high importance Barfield placed on user acceptance. “We were upfront about the sacrifices users would have to make on a daily basis and what benefits we hoped to gain. We also had users fill out a perception survey both before and after the trial. This gave us the chance to address users' fears about how their fingerprints would be controlled by explaining that there was no record of their fingerprints and that the fingerscan could not be used to reconstruct their forensic-quality fingerprints. After the trial, most users believed that biometrics had increased overall security.”

While Barfield believed the trial would be a success, he couldn't take the chance of compromising security at the trial site. “We didn't want to make ourselves vulnerable if the system didn't perform as expected,” he says. “That's why we chose a site which had a turnstile and guard who could punch people through if needed. We also left the existing wiring in place so we could swap back the old system in minutes if the biometrics system failed for any reason.”

FOR THE RECORD…

About the author

Ram Sathappan is biometrics solutions manager for Texas Instruments, Dallas.

About the companies

For information, circle the Reader Service number (listed below) or visit securitysolutions.com

Bioscrypt	15
MDI Security Systems	16
Texas Instruments	17

Want to use this article? Click here for options!
© 2008 Penton Media Inc.

This month in Access Control

• Opening Up About Door Closers
• An Enterprise Approach
• The Framework For Open Systems
• On A Higher Plane
• More from April's issue