Army builds major biometrics repository

The Army’s Project Manager for DOD Biometrics, located in the Program Executive Office for Enterprise Information Systems, is the Defense Department’s executive agent and lead agency for the development of all biometric efforts. Heading that work is Col. Ted Jennings, project manager of DOD biometrics.

Jennings and two of his chief deputies — Maggie Patton, product director for tactical biometric systems, and Greg Fritz, product director for biometric enterprise core capability — spoke recently with Defense Systems contributing editor Barry Rosenberg about their efforts to build a central repository of the DOD and methods being used to track who comes and goes in Iraq and Afghanistan.

DS: What do you see as the primary role of PM DOD Biometrics?

Jennings: We’re addressing two different product lines. Greg Fritz’s team manages the enterprise level and the central repository. So if you think of the FBI’s main database for all their biometric data, Fritz’s team developed a system that was based on their prototype, and we’ve taken it to places that just don’t exist anywhere else. We intend to put that forward as a formal program of record in both the acquisition system and the budget system. The other half of that is the data that feeds into Fritz’s system, and that is in Maggie Patton’s portfolio. She has all the tactical collection devices that are out there.

Both product directors are working to ensure that we’re in line with DOD and presidential directives to include what the other services might be doing. In particular, Patton’s group is dealing with the formal program of record that is supposed to incorporate or work with all of the different services’ biometrics efforts.

DS: What sort of biometrics data is being collected presently?

Patton: Our primary modalities are a full 10 print with slaps and rolls, which is when you put the upper part of your hand down and get all four fingers down to about the second knuckle. We also have at least one system that does the lower palm, the upper palm, and what is called the writer’s edge, which is the side of your hand away from your thumb. All of those areas of your hand have unique identifying prints on them. The more information we can gather, the likelier chance that our folks who do latent captures from IED fragments or sensitive site exploitation will be able to match against that latent print we pick up.

We also capture a facial array and the two iris patterns. We do have some places where DNA is being captured, as well as voice.

DS: How do you determine whom you capture this information from?

Patton: All foreign nationals, including local nationals and third-country nationals, who are gaining access to our bases. At major bases in Iraq, we’re capturing 100 percent of the population. Other places, it may be spottier. When the Marines went into Fallujah several years ago when it was a hotbed, they basically rolled up everybody in the whole city. From that point forward, when somebody was coming in or out of town, they could say "Do you belong here, or do you not belong here?"

DS: What’s the process when someone wants to enter a base?

Patton: For anybody who is on a long-term contract, the process starts with the employer, filling out a paper application, knowing what privileges they should have and whether they should be escorted or not. That goes into an automated system along with the person’s biometric information and gets vetted against our databases in West Virginia, which includes the FBI database. We have had folks wanting access to our bases in theater pop up with a criminal record in the FBI files. So then that information — hit [or] no hit — gets fed back to somebody at the local base, because every local commander gets to decide who can or can’t come on his base. And based on that information, whether it’s innocuous or whether they think it shows a threat, as well as other sources of information, they make a decision whether to allow that person on base. If they say “yes,” then we badge them with a card that looks a lot like a [Common Access Card].

Biometrics can’t tell me who you are, but it can tell me if I’ve seen you before. And based on those previous encounters, I’ve got more information to make the decision about you. Do I let you on base? Do I let you in the town? Do I capture you? So to get back to your question about base access, when the guy shows up at the base, I can check to make sure his card is a good card, and then I check your biometrics against the biometrics stored on the card so that I know you’re the guy I vetted and not somebody else using these credentials. Each time we encounter the person, we can verify that they are who we have dealt with before.

Jennings: You have to understand that one of the key tenets to [counterinsurgency] is working with the people. For instance, we want to help the people by not causing them to tell us where the bad guys are, or tell us if they think that they’re being infiltrated, because that then puts them in a position where the bad guys are looking for the people who are telling on them.

So we have now given them a tool to help secure their area. If we encounter somebody, we’re the ones determining based on our own information whether that person should be there. It also helps tell us where the good people are, and we want to work with the good people.

Patton: It allows us to employ people in the local population, that is, to help get the economy stimulated and to make sure people are gainfully employed instead of placing an [improvised explosive device] somewhere because somebody is willing to pay them $100 and they can’t do anything else. At the same time, it gives us a level of assurance that we’re not letting bad guys on bases.

DS: But you never know when a good guy becomes a bad guy, such as what happened with the Army psychologist at Fort Hood.

Patton: Right. That is why our database in West Virginia is continually vetted against any new information. Yesterday, I may have decided you were a good guy and given you a badge to a base, but tomorrow, if a latent print comes in from an IED fragment that matches against your fingerprints, we’re going to immediately send that information back to that base so they can revoke your credential, and, hopefully, when you show up for work tomorrow, invite you to stay.

DS: Are these mainly commercial-off-the-shelf technologies that you’re employing?

Jennings: It’s commercial off the shelf, but I have to tell you that in the area of collection of data, Fritz’s team has woven together the commercial technology in the back end and central repository in a way that nobody else has. Without his team’s contribution, then all the collections in the world are not taken to the level that they need to be to make sure we have the right information.

DS: Let’s talk about the back end. How have you customized for military purposes something that the FBI has developed for domestic purposes?

Fritz: It’s interesting that you should mention it because the foundation that we built our DOD Automated Fingerprint Identification System (AFIS) upon was the FBI’s Integrated AFIS from about five or six years ago. The Justice Department and DOD are pretty much on parallel paths right now, but there are differences.

In DOD AFIS, what we’ve done is take commercial technology using IBM blade servers, Cisco routers and switches, Oracle database, NetApps storage-area network, and a variety of other commercial products such as Linux and Microsoft workstations. L1 is currently the algorithm that we use, which is a commercial algorithm that does the matching. We’ve integrated it all to provide multimodal matching in the four modalities that Patton described: face, finger, iris and palm.

On top of that, we have a fusion algorithm, which provides a mathematical way to take scores in each modality and helps us get away from what we call “yellow resolves,” which is an inconclusive match. The algorithms are designed to produce automatic nonidentifications, which tell us "no matches here," automatic identifications, which tell us “this is definitely a match.” And the yellow space in between the automatic nondentifications and the automatic identifications is called a yellow resolve. With the fusion algorithm, we were able to reduce that yellow resolve window by a factor of approximately three.

DS: With that improvement, are you able to then quantify that back into catching bombers or keeping bad guys off bases?

Fritz: Yes, we are. We’ve identified increase in matches not only in the new modalities that we introduced but also in the fingerprint modalities. So we increased the incidence of automatic identification on the fingerprints and added face, iris and palms to the FBI’s AFIS software, which is just for fingerprints.

Increasing the amount of automatic identifications frees up the biometrics examiners so that they can focus on the really hard ones. Their workload is reduced because the algorithms are doing all the heavy lifting and making a lot more automatic nonidentifications or automatic identifications matches.

DS: Too much information with not enough tagging and retrieval capabilities is a problem in many areas, such as in the collection and dissemination of unmanned aerial vehicle sensor data. What about all the biometrics data you’re collecting?

Fritz: I would say that it is a challenge. Right now our database is measured in the millions. As we look to the future, we’re on a glide path to build that database out into the tens of millions and to increase our throughput from approximately 8,000 submissions per day to more than 40,000 submissions per day, while building out the database by a factor of approximately eight in the next two or three years.

DS: What are the technical challenges related to the collection, storage and retrieval of that data?

Fritz: Our challenge is getting all the information that Patton can provide. I don’t mean to speak for her, but a large part of her problem is that last tactical mile. She’s got small pipes to manage all of her collections and then to forward them. By the time they get to my system, that is not so much a problem.

I’m referring back to your question about information overload. Where I sit is almost sanctuary compared to being out on the battlefield. My system is nestled away in the mountains of West Virginia with big pipes coming in, and we can take just about everything that the tactical world can handle. My challenge will be when we build out the infrastructure to that last tactical mile and all of those sensors are able to feed the enterprise system in real time…then I’ll have a challenge. And that is why we’re already building out by a factor of two and then four in anticipation of increased submissions per day.

There is nothing but an upward trajectory in everything related to biometrics. There is going to be more records, and they’re going to be coming at us faster and in higher resolution and with more modalities. The database is going to grow in leaps and bounds as technology continues to mature.