03 September 2019

Ambitious projects for US’s defence- MARS and JEDI

Liviu Ioniţă

The Defence Intelligence Agency (DIA) announced that in order to accomplish their main purpose, which is giving their military personnel quality information, they consider for their future missions to be… MARS, a data warehouse dedicated to replace the current database.

Image source: Mediafax

Defence Intelligence Agency’s “space” mission

They are not talking about solar system’s planet, but the Machine-Assisted Analysis Rapid-Repository System (MARS). Concretely, MARS is a data warehouse, created to replace the actual database, Modernized Intelligence Database (MIBD), projected back in 1996, to work with a data storage space necessary for the military intelligence, to be complex, flexible, rigorous, scalable, to match the current century, based on cloud computing and machine learning.

The data storage is a term which refers to a collection of subject oriented data, integrated from many operational databases, non-volatile and time varying, dedicated to offer support when taking a decision.

The architecture of a data storage involves the description of the data and metadata which are stored in a monitoring and administration repository and toolsets.

The role of such data storages is stocking the data taken from the rest of the intelligence systems, in order to analyze them through different methods.

The cloud storage refers to organizing the data kept somewhere in the virtual space, wherefrom it can be accessed by any user, based on access.

The technology involves a set of complex IT services, on many servers placed in unknown locations for the user and offers significant advantages regarding the storage space, data access from any place, computing speed and data confidentiality and security.

The machine learning, a subset for the artificial intelligence (AI), are able to operate a great data volume, based on mathematical algorithms, to learn by themselves and upgrade their actions. The systems based on machine learning are controlled by humans, within the parameters established by those who are developing them.

This technology seems to be a promising approach when it comes to databases, as it allows the data system analysis, as well as system’s automatization and update. This can ensure a better data security. For example, by learning how a virus acts, the machine learning algorithms can identify the pattern related to how these are accessing storage data, and the identified deviations are used to anticipate ulterior issues.

Companies as IBM, Microsoft, Amazon or Google are offering complex machine learning platforms. Rich Clayton, Oracle Analytics, states that “adding machine learning to a cloud eventually helps people organize their activity, build, train and implement data models…”.

The current system, MIDB, used by all US, Australia, Canada and United Kingdom defence structures to integrate different types of data (text, images and maps) in a accessible representation  for analyst, no longer answer to initial demands, and is considered to be old-fashioned, centered on the human, stiff, inappropriate for a world mostly based on data.

Consequently, the Intelligence Defence Agency needs a new technological resource, dedicate to ensure quality’s increase and the effectiveness of intelligence products establishment process aimed at supporting the operations and offering strategic and early warning solutions, including the international partners.

According to Terry Busch, head of Integrated Analysis and Methodologies division, as it comes from cloud computing developments, artificial intelligence and machine learning, MARS will have greater capacities than the current databases, expected to create a virtual model and make analysis.

According to Defence Intelligence Agency’s documents, MARS will provide a virtual military intelligence environment for both task force and analysts, similar to what the World Wide Web represents for the common Internet user.

MARS will be more than a static product, it will not only solve problems, in an agile way, but its capabilities will continuously improve as it matures.

Currently, the Agency relies on physical servers spread on various networks and the manual data introduction, so that even with a four times larger number of analysts, the amount of data to be analyzed would still be overwhelming. The new cloud-based system will automatize most of what intelligence analysts do now (Tom Dillaplain, MARS project manager).

In order to be effective, the US Government plans on using extensive automatization and advanced analytical techniques for a wide range of data, information and knowledges. MARS will make the transition from the validation of an intelligence analyst’s work to the validation of an algorithm that can produce intelligence from the raw data. The system will also allow action courses’ simulation, giving operators the opportunity to quickly and fully understand the likely effects of the proposed activities and operations.

Terry Busch states that if 2019 was about learning, it was the training year, in 2020 they will move quickly from design to construction, so that by the summer of next year, MARS’s interface will be known. Also, some of its components will be available to US’s closest allies, so leaders in these countries can begin to analyze how they can use this data (Irving Townsend, data management officer).

Small obstacles for these big projects

But the project that is intended to be a large one and to make a change in how the intelligence community transforms data into military intelligence, faces some challenges that must find a solution before becoming operational.

Therefore, the Agency will have to make sure that data entry in the system will be done consistently, which is not that easy. According to Terry Busch, there are 1,300 different data standards within the Defence Department. Data standardization and interoperability of data collected by different agencies and services is essential for MARS.

Then, there are definitely some drawbacks related to storage. The amount of data collected by the information community that MARS must include is huge. For example, storing images and video files takes up a lot of space. So MARS will not store all the data itself, but will refer to the data on other agencies’ servers and, consequently, it must be able to index this information.

Another problem is analysts’ availability, from various the national security system’s structures, to share their work with each other and how will the access to information from another system be made, to how they got the data that’s at that information’s base. An impediment is, thus, the intelligence agencies and contractors who do not want to disclose their data collection methods or their own information.

While MARS will incorporate all MIDB data, some older systems will not be able to use the new system and will remain dependent on MIDB. This will make the transition difficult, and the Congress may not like the simultaneous funding of both programs (Terry Busch).

For General Robert Ashley, the director of the Defence Intelligence Agency, the biggest problem the agency faces can be summed up in one word: interoperability, the ability to transfer information between systems and partners.

For Ashley, sharing information with other members of the intelligence community and services is essential. And if processing speed issues could still be solved, the challenge is only being able cooperate effectively with other structures or other states.

It is important for the information to be approachable to more people, and agencies, including the Defence Intelligence Agency, must do whatever it takes to do it. And, according to Robert Ashley, a first step would be removing the term NOFORN (Not Releasable to Foreign Nationals / Governments / Non-US Citizens) from materials, which, comparing to RELTO (releasable to), implies that only local authorities are authorized to see the material, which not available for the allies or coalition partners.

MARS should be able to overcome this obstacle, as it should also remove those stovepipe systems, a derogatory term used for a system that can share data or functionality with other systems, but which does not actually do it.

This is the exact same reason why it is necessary to revise and modernize the Joint Worldwide Intelligence Communications System, that information community’s Intranet that analysts rely on to send classified information. Furthermore, the intelligence community must have some clearly defined domain boundaries and defined final flows.

A $10 billion JEDI for the Pentagon

And, as if they weren't enough to regulate, in the meantime, the Pentagon was facing a new challenge emerged due to the Joint Enterprise Defense Infrastructure, whose acronym, JEDI, also makes us thinks of ... the stars.

It's all about switching the Pentagon to cloud computing and replacing outdated, segmented systems, after a November last year check found that systemic flaws within the Defence Department networks are encouraging hacking, and finance systems were so disorganized that they couldn't be audited.

The Defence Department is asking cloud providers to build a secure, high-availability storage environment that supports the quick development and deployment of any application. Pentagon leaders want JEDI to include mobile servers, even miniaturized, also able to provide tactical units with classified, critical information for missions and action.

The JEDI project started in March 2018, when the tender call was launched, and Amazon, Microsoft, Oracle, IBM and Google were among the companies that have showed their intention to take over the $ 10 billion contract.

It seems that Amazon was the favorite company, probably due to the fact that its division, Amazon Web Services, already dominates the American cloud computing. AWS stores classified data for the CIA, supports federal agencies, including the Justice Department and NASA, manages the national immigration cases management system and stores millions of IDs.

This, for Amazon, only means that members of the defence, intelligence and national security communities deserve access to the best technology in the world.

The competing companies do not share the same opinion

Google dropped the JEDI contract after employees complained that the project did not fit company's culture, raising questions on the ethics of providing technology to the military field.

IBM has filed a complaint with the Government Accountability Office, and Oracle has taken the matter to court, both companies claiming it is a contract designed to favor Amazon.

Consequences emerged immediately. The Pentagon stopped the contract awarding process, which was about to be ended in September, as tensions has increased amid claims by President Donald Trump that Amazon was involved in a conspiracy. There is an older conflict between the White House leader and the company, as Trump claimed, among other things, that The Washington Post (owned by Amazon CEO, Jeff Bezos) is a Russian asset.

Given this context context, the Defence Department and the Defence Intelligence Agency seem to have a long way ahead until accomplishing what Robert Ashley called a moonshot, and the ambitious MARS and JEDI will actually become operational.

Most likely, Jack Gumtow, Agency's Chief Information Officer, was considering not only the coalition partners when he stated that they need a mindset change and proactive collaboration to achieve these goals.

And, the MARS and JEDI “moonshots” to be more than just illusions, as one of this word’s translation, but become feasible, necessary for the intelligence field.

Translated by Andreea Soare