Intelligence Architecture Augments Area Expertise With Data Access
Being able to extract useful information from archives is the new vital talent for rapidly changing operations.
The U.S. defense intelligence community is changing its information philosophy from emphasis on-call functional or geopolitical expertise toward rapid access to relevant knowledge from vast data files. To accommodate this shift, new technologies are enabling planners to implement an information architecture designed to provide authorized customers with streamlined access to vital information or expertise.
This change owes its origins both to the end of the Cold War and to technological capabilities introduced by the information revolution. Traditionally, the human in the loop in intelligence analysis was an expert schooled in a function or a particular region of the world that was considered a likely operations site or adversary. This approach served Cold War needs, where the Soviet bloc challenged the Free World in a number of specific theaters. Similarly, the monolithic bloc served as a clear intelligence target and organization nexus for the community.
Now, however, Western military forces frequently find themselves deployed for operations, ranging from those other than war to coalition conflict, in unexpected regions. Experts for these unfamiliar areas are few and far between, and analysts must rapidly pull together diverse data to generate an accurate picture of each region.
So that defense intelligence customers can access the relevant data they need to generate an accurate intelligence picture, the Defense Intelligence Agency (DIA) is developing the joint intelligence virtual architecture, or JIVA. The program aims to link intelligence users with their data rapidly and efficiently over a vast network. The JIVA effort features early involvement by allies that traditionally share intelligence with the United States.
JIVA would change intelligence dissemination from a "push-or-pull" methodology to a networked environment where an authorized user can extract needed information seamlessly from diverse data sources. This information would be updated as soon as analysts generate a new product relevant to the user's needs. It would effectively operate as an intelligence Internet, but implementing it also will require a major shift in the way data is processed and generated.
Col. Reid S. Huff, USA, JIVA program manager, explains that the new architecture is designed to change the way defense intelligence is analyzed, produced and disseminated. The program focuses on 14 command service intelligence centers, but it also reaches down to dissemination to a diverse customer set. Its driving functional requirements encompass approximately 12 areas, covering capabilities, methods and procedures. These capabilities will be introduced evolutionarily in five-year increments over the next 10 years.
Col. Huff notes that today's technologies could meet targets set for 2005. Many of these are either in prototype form or about to be offered for contract. "It is literally there for the taking," he emphasizes. "The harder difficulty is how to bring it into my work process." This promises to involve changes in processes, procedures and the work force itself.
"The technology for us is the Internet," the colonel continues. JIVA is designed to involve the Intelink and multiple levels of classification. The difference is the notion that the community no longer focuses on platform-centric computing. Instead, mid-tier computing and the Internet/intranet predominate. The system's basis for interoperability is a Web browser. The result of these innovations is an architecture that can be scaled for multiple users and crisis surges, Col. Huff states.
"The nature of our work force now is that the issues we are dealing with are so transitory, we are constantly having to shift and manage our work force, even though we have specialization," he notes. "What we need to do is provide them with tools that they can carry with them across problems. [This would permit them] to access mapping data and move it around the world because you never know what you are going to need next."
JIVA's intelligence product effectively is a Web portal through which a user navigates to obtain the desired information. An operator seeking information on a topic would activate hot buttons on a display that open Web linkages, or pointers, to a distributed knowledge base.
For example, a user seeking information on a region would touch the target site and navigate through different types of data. The operator could learn about airfields by viewing actual images, reviewing runway information, and scanning terrain data, to name a few options. This information would be drawn seamlessly from a number of diverse sources.
To enable this capability, JIVA requires that intelligence knowledge be separated into modular components, or objects, that can be stored and retrieved according to their content. This object design will ultimately supplant the reliance on relational databases and support the dynamic creation of living intelligence products. This modular reuseable intelligence approach also helps shift the focus away from stovepipe products to a process that allows for the rapid recombination of intelligence produced by multiple agencies.
The driving force behind the concept is metadata, the colonel declares. To generate this material, an analyst enters the information. This information is tagged by content, structure and classification. For dissemination to various classification levels, including foreign intelligence services, the information is electronically wrapped in top secret, secret, commonwealth, North Atlantic Treaty Organization (NATO) and other allied versions. This write-once, distribute-many approach also allows automatic updating when new information supercedes older data. The system's metadata catalogs constantly pull information into a topic.
A concurrent effort involves rapidly automating the information flow process. It currently can take several months to generate a full transportation study on a country to which U.S. forces may be imminently deployed. Bottlenecks such as editing, release and disclosure contribute to the delay. Automating workflow is the key to streamlining the process, the colonel acknowledges.
The system's content-tagging approach also will help eliminate ambiguities. Seeking information on a typhoon, for example, could generate responses based on tropical storms or Russian ballistic missile submarines. Under JIVA's Internet language methodology, this dynamic information bank would be authored, stored and retrieved based on content and precise language.
Existing relational intelligence databases could contain information on a facility such as a tank factory, for example. When the tank factory is converted into a baby carriage factory, as was one case in the former Soviet Union, the only way an analyst would know of that change is by reading specific message traffic. The different forms of data hinder them from being collected into a single, useful domain. With JIVA in place, this new information would be accessed by the user wanting to know about the tank factory. Users would not have to wait for scheduled updates to learn the latest information about their target of interest.
Providing this type of information surge capability addresses one challenge facing the architecture designers. Another is giving the typical desk analyst digital access to information easily and rapidly. Solving these challenges requires changing intelligence production to a common format to provide broad-based access to the mountains of data currently being generated that are increasingly difficult to retrieve.
"We need to be able to move information as objects," Col. Huff declares. "We can't do that today because we produce big documents of information rather than information."
Stovepipe and other legacy systems are only some of the tasks facing the defense intelligence community as it retools for the information age. Traditionally, the community could invest in infrastructure, but it ran into difficulties reaching departmentwide consensus on fielding capabilities.
To overcome this problem, Col. Huff notes that the JIVA program consolidates the relevant financial authorities of the Defense Intelligence Agency with the U.S. Army, U.S. Navy and U.S. Air Force intelligence organizations. Service executive agents are funded during years of execution, which allows considerable flexibility to build and integrate capabilities, he says.
As designed, JIVA comprises four components: enterprise services, the analytic environment, production and dissemination. Ultimately, collection management may be added, the colonel notes. This reflects the interactive nature of the components, which could permit reuse of some technologies.
Enterprise services feature collaboration, which encompasses bringing users from multiple sites together in a common virtual meeting place to share information. This draws from commercial developments that saves the military research and development costs. General Dynamics' Information Workspace currently is fielded in all joint worldwide intelligence communications systems sites and at test sites on the secure Internet protocol router network in three commands. It also is fielded with several commonwealth countries allied with the United States. Another enterprise service, desktop training, uses collaboration as its delivery system for interactive learning.
The goal in enterprise data archiving is to allow for message traffic, open source information, structured data stores and proprietary information, which would be accessed through content tagging. Currently, intelligence and messages are managed on a site-by-site approach based on a statement of intelligence interests. This approach to data tends to reinforce stovepipe thinking, Col. Huff explains. Under JIVA, every analyst, regardless of location, would have common access to a complete dataset and the ability to manipulate this information using a universally accessible tool set. This avoids the problem of different sites having unrelated tool sets and network topologies.
"What we want to do is to broker access independent of the design of these [different] databases, perform security authentication, and audit who is taking what out," the colonel says. This lets different forms of data effectively be converted into a common framework, where they can be visualized in the same fashion.
The JIVA analytic data environment (JADE) will begin with an enterprise version of the Broadsword gatekeeper, which will be fielded defensewide in July by its developer, the U.S. Air Force 497th intelligence group. This version will support XML tagging of query returns and allow comparisons of dissimilar data types.
Another JIVA element being fielded in July is the Pathfinder analytic tool. This will give an analyst 31 automated techniques to support data understanding, including relationships, linkages, dependencies, timelines and visualizations. Volumes of text data amassed over several years can be reduced with a single query for display and annotation. The JIVA enterprise version of Pathfinder, which is integrated with JADE, has been converted for Web use and will be fielded defensewide by its developer, the Army National Ground Intelligence Center.
This tool is largely geared toward power users, who typically understand information technology. They tend not to be rank-and-file analysts, however, so the JIVA program may incorporate more simple versions of off-the-shelf capabilities, Col. Huff adds.
The importance of geospatial data in intelligence operations has its effect on JIVA. The National Imagery and Mapping Agency is teaming with the DIA on building a prototype system based on commercial technology. The colonel explains that this effort comprises a set of terabyte servers that host geospatial information and are accessible using Web technology. This prototype is scheduled to be ready in 2000, with initial operation slated for 2001.
Within the program, work on digital production began with a September 1999 contract awarded to BTG for the company's InVision product. This lets a desk analyst generate a multimedia product in Internet language. Col. Huff explains that InVision "protects" an analyst from the technology by operating transparently according to his or her word processing. Analysts can collaboratively assemble intelligence documents from reuseable component modules and establish or discover interrelationships.
The production prototype is running at the U.S. Pacific Command. Evolutionary beta versions are scheduled for deployment at four major intelligence sites this year. The authoring environment would be fielded to all production centers by 2001. The operational production version would come on line in the 2002 to 2003 timeframe. Its introduction will represent a fundamental redesign of the intelligence community's business process, a change in work force training, and possibly new position reclassifications, Col. Huff declares.
Proper security in this intelligence network is achieved through a commercial lightweight directory access protocol server with embedded public key infrastructure X.509. For example, access for a user attempting to enter a virtual meeting space would be confirmed by a security certificate that is online within the registry. This is not the final configuration, however. Col. Huff suggests that program developers are looking for remedial solutions this summer to enable them to buy "a real system that makes this happen."
Until JIVA has a multilevel security capability, users would download different versions of each document. Multidomain dissemination "is considered by most of the agencies as the route that we probably can take" to achieve multilevel security, the colonel says, and it is facilitated by this type of system.
The program administrators purchase applications that are Web-centric and platform independent. Licenses are acquired on an enterprise basis, as opposed to seat costs. Traditionally the domain of major components from firms such as Oracle and Sybase, this approach soon will reach down to items such as collaborative tools, the colonel offers.
Advanced Web tools are key technology enablers. The colonel notes that the commercial sector is developing new search engines that he describes as phenomenal. These capabilities will be coming into the marketplace soon.
Among hardware, a revolution is dawning in high rates of input/output. This will allow the rapid reconfiguration of a multidomain server. New network monitoring tools also will permit just-in-time upgrades to servers.
The biggest challenge to JIVA implementation is its groundbreaking nature, Col. Huff shares. "We are dealing with capabilities for which there are no performance metrics," he relates. "We are building to satisfy a work force behavior that does not exist. And yet, unless we put it out there in some form, we will never attain that." The program is trying to strike a balance between easing into the new environment and "coming in like gangbusters," he notes.
A related challenge is to stay ahead of the technology without reaching too far. A company may have developed a similar type of system for the commercial or civil government marketplace, but its lessons may not apply to a global intelligence system. Even the JIVA prototype elements under evaluation at command sites do not necessarily translate their newly learned lessons to a global scale.
JIVA designers also are trying to strike a balance between a structured database configuration and information anarchy. Early prototyping produced a configuration that was too structured, so the program is learning from that experience and is loosening its bounds. "If we create something in this process so rigid that it looks like a relational database, then we have failed," Col. Huff declares.
On the other hand, planners must deal with the potential for JIVA to overwhelm itself and its users with complexity. For example, a piece of information for simplicity's sake could be tagged only by its title. At the other extreme, every single word in a report could be tagged, which would create chaos. Users would no longer have the ability to perform tailored searches. The colonel admits that the technology is too new for the correct answer to be clear cut. "We have some real decisions about how far to go," he maintains.