National AI Research Resource Task Force seeks more ‘diverse’ public responses

The task force designing a blueprint for a National Artificial Intelligence Research Resource extended the deadline to publicly comment on its work by a month to elicit more diverse responses.

A few expected respondents asked for more time, citing the depth of questions in the request for information (RFI) and summer vacation conflicts. So the deadline was moved to Oct. 1, according to a Federal Register notice posted Wednesday.

The 2021 National Defense Authorization Act directed the task force to create a roadmap for a sustainable resource that equitably meets the computational and privacy-preserving data needs of AI researchers.

“Given that we want to hear from as many folks as possible — we’d really like a broad and diverse set of perspectives among the responses — we felt it best to extend the RFI response date by a month,” Erwin Gianchandani, who co-chairs the task force as the National Science Foundation‘s representative, told FedScoop. “We believe that timeline will maximize responses while still allowing us to digest the responses and use them to inform the task force’s discussions and deliberations.”

The NAIRR blueprint is expected to outline campus, cloud, high-performance and supercomputing assets; data and data governance; educational tools and services; user interfaces; and privacy — all while ensuring researcher access regardless of gender, background, expertise or any other demographic.

By giving academics and startups, in particular, more time to respond to the RFI, the task force is ensuring that the diverse array of researchers it wants using NAIRR inform that blueprint.

“It’s not really about the big tech companies that already have lots of computational infrastructure and data,” said Lynne Parker, the other co-chair and White House Office of Science and Technology Policy representative. “I think it’s fairly easy to agree that this resource is not intended for them, even though they may indeed have some great ideas and are absolutely willing to engage to help in the design of a good resource.”

Google and IBM are still two of the four companies representing industry on the task force, along with four universities and four agencies that also include the Department of Energy and National Institute for Standards and Technology.

The task force held its first of four public meetings this year in July, with the next one scheduled for Aug. 30 and four more slated for next year. In between working groups will meet for deeper dives into specific aspects of the NAIRR blueprint.

An interim report is due May 2022 and the blueprint November 2022 to the president and Congress, though it will fall to others to actually implement the plan.

The blueprint must be scalable, easy to use, affordable — as in not cost billions of dollars — and sustainable.

“It [has to be] adaptable, in that we’re able to add new resources or keep the computational resources up to date,” Parker said. “Maybe there’s some new kinds of computation, maybe a new AI chip that we want to provide lots of people access to use and explore their research ideas on, and so we want to be able to have the resource be adaptable and have these new, different kinds of architectures.”

Census Bureau missed opportunities to avert 2020 cyberattack — report

The U.S. Census Bureau was slow to report a major cyber breach in early 2020 and missed key opportunities to limit vulnerabilities, according to the Department of Commerce’s Office of Inspector General.

An investigation published Wednesday by the oversight body found that the bureau was failing to scan its remote server prior to the cyberattack, and the failure to keep sufficient system logs hindered a subsequent investigation into the incident.

“We found that the bureau should make improvements to its cyber incident response process. Specifically, the bureau missed opportunities to mitigate a critical vulnerability, which resulted in the exploitation of vital servers,” the IG said in its report. “Once the servers had been exploited, the bureau did not discover and report the incident in a timely manner.”

The breach took place in January 2020, when hackers accessed the agency’s remote servers. None of the compromised servers were involved with the 2020 census, and the count was unaffected by the attack.

According to the report, the bureau was also operating servers that were no longer supported by the relevant vendor.

The IG has issued a string of recommendations, including that the Census Bureau’s CIO review procedures for notifying IT staff when critical vulnerabilities are publicly disclosed. These include the recommendation that the CIO review the automated alert capabilities of the bureau’s security information and event management tool.

The IG also called on the bureau’s director to ensure that the CIO incorporates periodic reviews of IT asset configurations and also updates bureau response policies to include a specific timeframe.

Earlier this year, FedScoop revealed that the Census Bureau had installed Luis Cano as its chief information officer after previously working as chief of the agency’s Decennial Contract Execution Office. He replaced Kevin Smith, who left the bureau in January to join the Federal Housing Finance Agency.

Under federal policy, the statistical agency is required to adhere to the Department of Homeland Security’s Continuous Diagnostics and Mitigation Program, which operates regular vulnerability scanning of all systems.

Jay Bonci sworn in as Air Force CTO

Jay Bonci was sworn in as chief technology officer within the Office of the Air Force CIO on Tuesday.

He is a former long-time employee of cybersecurity and cloud service company Akamai Technologies, where he held senior roles within the company’s Department of Defense and federal tech divisions, including as senior director for public sector engineering.

Bonci will oversee a vast tech portfolio in the Air and Space Forces within the office of current CIO Lauren Knausenberger. The CTO’s primary duty is to “advance the Information Technology landscape,” according to the Air Force website. He takes over from Frank Konieczny, who in February left the role to take up a private-sector post.

“My part of the mission is to empower and accelerate those on the ground by providing a cohesive enterprise architecture and service delivery strategy,” Bonci wrote on LinkedIn Tuesday.

The Air Force has two other CTOs: one at the Air Force Research Lab that focuses on emerging tech, and a science and technology CTO created in 2019 that focuses on “disruptive capabilities.” Bonci’s role is more cyber- and IT-focused, the Air Force has previously said.

Bonci added in a LinkedIn post that his priority will be developing and integrating technologies, up-skilling the workforce and monitoring investments in the tech budget.

“I hope my contributions demonstrate a massive impact in a short amount of time,” he wrote.

Network modernization strategies to support modern applications

Cloud driving ‘sea change’ in federal data utility, says State’s cloud director

The advent and ubiquity of cloud computing in the federal government has largely increased the focus of federal officials — particularly senior leaders — in leveraging data as an asset, according to Brian Merrick, head of cloud programs at the State Department.

In an interview with FedScoop, Merrick said there has recently been “a sea change” in the way officials think about data, due in large part to the shift in models from data tied to an application to one with a more openly architectured environment where data can freely flow between applications.

“Our traditional sort of system-box diagram model that we’re used to in the old sort of waterfall application world, where your data is associated with an application and you have an owner of that application, everything’s tied to that and it’s about doing an interface between this box and that box. That kind of dynamic is no longer the driver,” Merrick said of data in legacy environments. “And so that frees up people to really look at that a little bit differently when it’s not necessarily tied directly to a specific application. Often what we’re finding now is there are many applications sharing data by virtue of necessity between these different environments.”

This has placed the focus on the data as the sort of lowest common denominator and most important asset in an IT environment. “It’s really about the data,” Merrick said.

On top of this, the cloud also introduced added benefits of data integration and sharing, speed of delivery, and evidence-based decision-making, he said.

“When a senior leader comes and says, ‘Look, I want to do this thing, I need an ability to get this particular task done,’ and we go back and say, ‘Well, you know what, if we have these two or three data elements from over here and combine it over here, we can get you that in a week,’ the doors open,” Merrick said. “And all of a sudden, they’re knocking on the door of the data owner saying, ‘OK, you need to make sure that this group gets access to these data elements, and we can fix this business problem.”

Such a structure has “elevated data concepts out of the IT silo to a senior management viewpoint,” Merrick said. “They now see the value in getting access to that data, which opens the doors to that data, really from a policy and an ownership standpoint, and I think that’s been a huge change.”

And as those senior leaders get more experience using data to add new value or make better decisions, they get hooked, Merrick told FedScoop.

“Leaders have realized that … they were probably not basing decisions in the past on as good of data sources they have now. And that has really increased the awareness of it and put the emphasis on it,” he said. “And I think we’re gonna see a lot more of that as time goes on. And as we start bringing in disparate data elements we hadn’t maybe thought about before, like sensor data from the Internet of Things — a whole host of other data elements from other outside data sources — it starts to paint pictures when we’re trying to solve real-world programmatic problems. And I think we will see that start to scale even faster.”

NARA targets digitization of 500 million pages by 2026 in draft strategic plan

The National Archives and Records Administration has published the first draft of a new strategic plan that includes an ambitious target for the agency to digitize 500 million pages of government records by 2026.

In its proposed 2022-2026 strategy, the administration said it hopes to process 85% of archival holdings by the end of the period. Additionally, it will work to enhance catalog descriptions that promote equity in discovery and archival access for underrepresented communities.

“In order to achieve success in this goal, NARA must digitize millions of records we hold in analog formats, keep pace with the continuous stream of new records we receive each year, and develop new ways to help citizens find our records through the online National Archives Catalog,” the agency said in its draft strategic plan.

NARA is seeking feedback on the proposed plan, which must be submitted by Aug. 20.

The agency faces a vast challenge in cataloging existing digital and physical records that have yet to be processed. One former senior agency leader speaking to FedScoop described “literally stacks” of wooden palettes containing hard drives that have yet to be processed at the administration.

NARA also intends to deliver 95% of customer requests — including FOIA requests — on time by 2026. It is formulating the new strategy amid sustained criticism over its handling of a recent veterans’ records backlog at its National Personnel Records Center division.

House lawmakers at the end of July wrote to NARA, calling it to apply for funding from the Technology Modernization Fund to help clear unprocessed records that tallied 500,000 in July. This followed a previous letter sent by a bipartisan group of politicians, asking for the Department of Defense to intervene.

AT&T expands IT apprenticeship program supporting national security agencies

AT&T is expanding an IT apprenticeship program that trains entry-level talent in support of federal national security agencies, the company announced Tuesday.

Northern Virginia Community College (NOVA) and the Virginia Department of Labor and Industry joined the AT&T Catapult program that sees students receive top secret-level security clearances for contractors while they’re trained.

The program helps meet a talent need with IT positions expected to increase at twice the rate of overall U.S. employment, and an estimated 11,000 jobs in need of filling every year in Northern Virginia alone, through 2030.

National security agencies including the Cybersecurity and Infrastructure Security Agency, as well as the Office of Management and Budget, have made strengthening the IT and cybersecurity talent pool a priority. This is especially true following the issuance of an executive order to boost diversity across government and OMB reaffirming its commitment to new talent pipelines for roles like cloud technology experts.

“Identifying and hiring qualified information technology talent continues to be a challenge for federal agencies,” said Jill Singer, vice president of defense and national security at AT&T Public Sector and FirstNet, in the announcement. “In particular, national security agencies face challenges such as high-level security clearances and a need for specific IT skill sets, which can make it even harder for them to grow their talent base.”

Students will train at NOVA’s Reston complex and work as part-time AT&T employees in person and virtually at AT&T’s Oakton facility.

Security clearances follow the standard vetting process for contract personnel and will likely take the two years of training to be processed, although students who receive theirs sooner will move to full-time employment at AT&T — with the understanding they must successfully complete their training to remain employed. AT&T declined to comment on which national security agencies were sponsoring clearances or benefitting from the influx of talent.

Program graduates earn 2,000 hours of on-the-job training in technical, soft and lab skills and obtain five industry certifications in A+, Network+, Security+, CCNA and ITIL.

AT&T Catapult began in Maryland at Howard County Community College in 2019 to share talent development costs with government, reduce churn, lower the cost of entry-level positions and limit the need for four-year degrees. NOVA and AT&T will access the success of the program annually based on participation and completion rates before adjusting or discontinuing.

Interested students have between Aug. 26 and Sept. 23 to apply for the program and will hear back by late October, before classes begin in January. Applicants must be at least 18 years old, U.S. citizens and Virginia residents, with other qualifications to be addressed during virtual, informational sessions on Aug. 26 and 30 and Sept. 1.

Northern Command using data platform to better detect incoming threats

The combatant command in charge of protecting the U.S. homeland has a new weapon in that fight: data.

A recent series of experiments hosted by U.S. Northern Command and the North American Aerospace Defense Command (NORAD) tested the military’s ability to use artificial intelligence to comb through data to detect incoming threats. The tech is aimed at increasing the precision with which the command can detect, and thus protect, against new kinds of missiles that can travel at multiples of the speed of sound, as well as smaller vehicles like drones.

“Data exists; what we are doing is making that data available,” Gen. Glen VanHerck said of the effort, during a press briefing on the Global Information Dominance Experiments (GIDE).

The command already uses lots of data, with analysts poring over sensor information to detect if intercontinental ballistic missiles from North Korea or Russia are headed toward U.S. cities. The experiments tested the command’s ability to use tech and available cloud infrastructure to expand how much data it can use and better integrate with other commands.

The Defense Department has been on a journey to enhance its use of data and artificial intelligence. The data-driven concept of operations behind that is Joint All Domain Command and Control (JADC2), a framework where all data can easily be shared from across air, land, sea, space and cyberspace activities, like a military Internet of Things. AI would be used on top of that to make sense of all the data and produce coordinated options for commanders.

“This is JADC2,” the general declared about the tests, called the Global Information Dominance Experiments (GIDE).

GIDE consisted of three iterations, each with their own focus areas but all broadly testing the ability to share data and use AI in homeland defense. The first GIDE experiment was a tabletop exercise testing how the NORTHCOM would collaborate with other commands if it had AI-enabled early warning systems. The second experiment introduced live tests of data sharing, with the third expanding those experiments and early warning systems.

NORAD and NORTHCOM “leveraged an interoperable data platform that is a cloud-agnostic, meaning the data integration software solution can work on whichever cloud capability is preferred,” a spokesperson told FedScoop in an email.

The experiments used existing infrastructure already in place in DOD. The idea was not to build or buy new tools but to leverage what the command already had available in new ways.

“GIDE’s success relied on existing capabilities used in innovative ways,” the spokesperson said.

The backbone of the tech is both already there and subject to change. The incoming Joint Warfighter Capability Cloud is being rolled out to meet the DOD’s enterprise cloud needs, providing compute and storage all the way out to the tactical edge.

“The GIDE 3 experiment showcased how the software tools designed for cross-combatant command collaboration, assessment, and decision-making can be used to enable more effective global logistics coordination, intelligence sharing and operations planning,” VanHerck said. “By integrating more information from a global network of sensors and sources, using the power of AI and machine-learning techniques to identify the important trends within the data, and making both current and predictive information available to commanders, NORAD and USNORTHCOM are giving leaders around the globe more time to make decisions and choose the best options available, whether in competition, crisis or conflict.”

NASA’s cloud push is motivated by data constraints

Editor’s Note: This story has been updated with information on the first missions to use the DAPHNE service, the clouds that are part of the project and a timeframe for mission launches.

NASA’s move to the cloud is in large part to accommodate the data it’s receiving from space and accelerate resulting innovations, rather than simply migrate applications to a more secure environment.

The agency operates satellite ground stations worldwide that are part of its IT footprint, and its Data Acquisition Processing and Handling Network Environment (DAPHNE) project aims to connect them to commercial clouds.

NASA has spent billions of dollars to date trying to address the growing challenge of transmitting data from next-generation satellites and overseas ground stations to make scientific discoveries, and the global scale of many cloud providers is hard to replicate.

“We’re going to be beaming back data from space and transporting it back to the United States via the cloud network,” said Joe Foster, cloud computing program manager at the Goddard Space Flight Center, during a recent ATARC event. “It’s saved us money, in terms of investing in infrastructure, and it’s allowed us to refocus those dollars into things like upgrading the actual satellite antennae themselves.”

The first missions to adopt the DAPHNE service will be the NASA-Indian Space Research Organization (ISRO) Synthetic Aperture Radar (NISAR), being codeveloped to monitor Earth hazards and environmental change, and the Plankton, Aerosol, Cloud, Ocean Ecosystem (PACE), a satellite that will help understand how the ocean and atmosphere exchange carbon dioxide. NISAR will produce 41 terabits of earth science data per day — which for comparison would take thousands of cellphones to hold — and PACE 4 terabits per day.

DAPHNE will allow that data to be processed at a fraction of the price, with cloud costs dependent upon each mission’s data volume. Exactly how missions will use DAPHNE depends on their launch schedule, which has shifted due to the COVID-19 pandemic.

NASA onboarded several commercial cloud services with which it’s growing operational maturity. Although DAPHNE currently only uses Amazon Web Services, it’s designed to be interoperable among various cloud service providers whose services will be used depending on mission requirements.

DAPHNE is part of the Near Space Network Initiative for Ka-Band Advancement (NIKA), which has a readiness review scheduled for the fall in advance of mission launches. Future missions that will also use DAPHNE include the Nancy Grace Roman Space Telescope, as well as the Spectro-Photometer for the History of the Universe, Epoch of Reionization, and Ices Explorer (SPHEREx).

NASA is also part of an open-source project, Pangeo, a managed Jupyter hub and data-science platform designed to be deployed in any cloud environment. Along with the National Oceanic and Atmospheric Administration and U.S. Geological Survey, the agency is using Pangeo to store machine-learning algorithms performing earth science.

In recent years, Goddard has placed legacy microscopes and other instruments with firmware too old to be upgraded in a Secure Lab Enclave, but data from those devices can still be analyzed in the cloud.

“We basically firewall off and put on a private network all of these legacy pieces of equipment that can’t get upgraded anymore,” Foster said. “That doesn’t mean that you can’t take telemetry data or files off of the piece of equipment and try to run diagnostics or things like that in a more modern cloud environment.”

NASA is standing up cloud-enabled projects at a faster rate because by departing from its old, contract-task order approach, which took as long as six months. Now projects are launched in 45 minutes, and the agency wants to replicate the National Geospatial-Intelligence Agency’s authority to operate (ATO)-in-day effort, where core services are approved at the enterprise level.

While NASA’s cloud platform is currently only accredited at a FISMA Moderate level because its science is largely public-facing, the agency is in the process of creating a FISMA High enclave for more sensitive data dealing with flight and launch capabilities and human space exploration. The enclave is part of NASA’s effort to implement the recent cybersecurity executive order and move from legacy, on-premise systems to a zero-trust, cloud-based model, and the Technology Modernization Fund could cover the cost.

“With the new executive order we wrote a proposal to the TMF [Board] to upgrade and create a FISMA High enclave,” Foster said.

Dataset demands and remote leadership among top challenges for CDOs during pandemic

The volume of demands for coronavirus-related datasets and communicating with staff in a remote environment were among the biggest challenges faced by chief data officers during the pandemic, C-suite advisers told FedScoop.

According to consultants and nonprofits canvassed, the responsibilities of CDOs expanded as the COVID-19 pandemic spread across the U.S. during early 2020, to include leading the flexible use of previously untapped data and working to establish a culture in which experimental data dashboards were used to help coordinate teleworking staff.

CDOs were also presented with the harder-to-grasp challenge of influencing department culture from behind their computer screen.

“[W]hile the role of the CDO was always important, the pandemic has highlighted just how important having a chief data officer managing and governing data across the federal agencies really is in practice,” said Nick Hart, president of the Data Foundation, a nonprofit that works closely with CDOs.

Hart continued: “[E]verything [during the pandemic] was happening at lightning speed, we were no longer in a position to wait a year for new data collection that could be linked to other government data sources, or even private-sector data sources.”

Agencies pulled together datasets at extraordinary speed to launch programs, such as the separate Department of Health and Human Services’ Protect Public Data Hub and Centers for Disease Control and Prevention COVID Data Tracker. Speaking at an event earlier this year, Kevin Duvall, acting chief data officer at the Department for Health and Human Services, described the challenge, saying the agency got “more comfortable” with datasets and the quality of data over time.

CDOs at pace had to consider a raft of ethical concerns, including how datasets might be used by lawmakers – and the way in which publication of such data might be viewed by teams internally.

“Collaboration between data visualization staff and software engineering teams really only happens if there is trust and knowledge that if any data is put out there [staff] will have the chance to validate it,” Grant Thornton public sector manager Tracy Jones told FedScoop.

“Other agencies I work at have very public-facing websites and public-facing data, and there are concerns on that front: Do agencies want to put data out there that constituents and policymakers at a higher level are going to be looking at?” Jones added.

Under a data ethics framework published in late 2020 as part of the Federal Data Strategy, agency data leaders are expected to understand and disclose any known limitations, defects, or biases — a challenge that CDOs had to embrace while working to respond to huge demands from their agency leaders and lawmakers.

According to Jones, one of the only ways to overcome internal concerns between teams over the publication of datasets is to foster a high level of communication. “You always want to understand: What is this data, how is it going to be used?”

In addition, the advocate role of the CDO — the ability to take staff on a journey towards the more effective deployment of data — also came under pressure during the pandemic, according to Jeff Lawton, Grant Thornton’s lead for enterprise information management solutions.

“I think that COVID has essentially hamstrung their ability to some degree, where they can’t use the full power of their charisma, their charm — because that’s part of the CDO position to get people to do something I want them to do that they are not normally doing,” he said.

On top of this, data leaders across federal agencies over the last 18 months have had to lead experiments with new datasets to help move staff to an all-telework environment during the height of the pandemic.

Speaking with FedScoop, one consultant described a rise in demands for more creative analytics from data leaders to afford more insight into exactly how staff were working outside the office. CDOs have had to lead the charge on the internal use of metrics, where it is efficient and makes sense within the teleworking environment.

“[We saw] a shift towards really trying to understand productivity…beyond maybe just the simple timecard entry. What are the ways that we can pull data to validate that we’ve all been productive in this engagement,” said JD Walter, a former HHS modernization adviser who is now executive vice president of solution optimization and execution at Golden Key Group.

Above all, however, Lawton added that one of the biggest determining factors in the success of a CDO — during the pandemic but also more widely in their role — has been attitude. He described the enthusiasm of one current, successful CIO as “infectious.”

“When we asked her what she wanted to have achieved in a year’s time, she said ‘I want people to have fun working on data governance,” Lawton said. “Who doesn’t want to be part of that group, which has that positive energy.”