DFAT is building its own facial recognition system

Australia’s passport authority is building in-house a new system to handle its facial recognition operations in preparation for the high availability that will be required from two separate government biometrics initiatives.

DFAT’s Australian Passport Office has been using facial recognition and other biometric technology since 2005 when it introduced Cognitec Systems software to help detect passport fraud.

It increased its use of biometrics in 2011 through the creation of a dedicated panel of providers.

But the impending arrival of a trusted digital identity solution being built by the Digital Transformation Agency, coupled with the new facial image matching hub run out of the Attorney-General’s Department (AGD), has prompted DFAT to take a long hard look at its own infrastructure.

The DTA’s ‘Govpass’ digital identity solution will provide a whole-of-government approach to verifying identity in online government transactions.

It sits between the government service and an identity provider - like DFAT or a bank - who confirms identity documents provided by the individual using the government service.

The solution is currently in beta.

The AGD-controlled national face verification service (FVS) lets law enforcement agencies share citizens’ facial images to verify identities, and identify unknown individuals, in order to reduce cross-border crime.

It was set live in November last year as part of a staged implementation, giving the AFP and DFAT access to images on citizenship applications held by the Immigration department.

Other types of images, like visa, passport, and potentially drivers license photos, and more government agencies including ASIO and Defence will be brought into the service over time. The AGD has said the government currently holds more than 100 million facial images for identity purposes.

The arrival of these two platforms has forced DFAT to review whether its own facial recognition system will be able to cope with the new demands.

DFAT processes more than 25 million facial images a year - around 8000 a day -  and that figure is growing at 2 million annually.

Stephen Gee, assistant secretary of the passport policy and integrity branch, says the agency has potentially the country’s largest database of biometric facial images.

To try and get more out of its facial recognition operations, and to cater for the “high levels of availability” Govpass and the FVS will require, the agency has embarked on a three-step "systems improvement" drive.

“The first is upgrading our architecture. At the moment we have single instance architecture with warm disaster recovery, which means we can get back on our feet within about an hour if the system goes down,” he told a Biometrics Institute conference in Sydney today.

“That’s fine for our current passport processing operation, and it’ll be fine for the new passport processing system that we’re developing and hope to put in place later this year.

“But the Attorney-General’s Department hub to which we’ll be connecting in July, and the DTA’s trusted digital identity framework initiative, will require high levels of availability. Being down for an hour won’t really be an option.”

To meet these demands and cater for expected future growth in the use of biometrics, the DTA is “building, not buying” a new facial recognition system, Gee revealed.

“It’ll have highly available architecture, including hardware stacks fully replicated across two sites and an F5 load balancer,” he said.

“It will be fully scalable and extensible as new uses are developed.”

Gee said development work on the new system was “well advanced” and it was on track to enter production in the first half of next year.

The department has also been quietly working with the Defence Science and Technology Group on “cutting edge” facial biometrics research intended to help develop standards for image comparisons and establish facial examination as a “recognised forensic science”, according to the agency.

Gee hinted that this research was a big component of DFAT’s systems improvement drive but declined to provide details.

“This is just a tease at this stage: we’re doing research with the DSTG about a new approach to facial recognition which we think might have the potential to revolutionise matching performance but I can’t really say much more about it at the moment,” Gee said.

He indicated plans to reveal more towards the end of the year “even if it is a failure” in order to aid wider development in the area.