Intelligence collection ontology

In the paradigm of intelligence collection ontology, which gives computer assistance to requesters of information and managers of information collection sources and methods, the requester is asked the question "What are the requirements of a mission?" These include the type of data to be collected (as distinct from the collection method), the priority of the request, and the need for clandestinity in collection. Collection system managers, are asked, in parallel, to specify the capabilities of their assets.

Outside the specific disciplines of intelligence, the semantic web is the most general area of research in these issues. Work in other disciplines, such as with the Semantic Network in the Unified Medical Language System® (UMLS®) may also provide ontological insights.

Selecting tactical imaging sensors

Preece's ontology is focused on imagery intelligence technical sensors, but also considers HUMINT, OSINT, and other possible methodologies.

The ontological model then compares the requirements of colletion against the characteristics and availability of collection resources. For example, in an example of matching a request for an unmanned aerial vehicle (UAV) to a mission, they define "the UAV concept encompasses kinds of UAV, which may range in cost from a few thousand dollars to tens of millions of dollars, and ranging in capability from Micro Air Vehicles (MAV) weighing less than one pound to aircraft weighing over 40,000 pounds...

• Small UAV (SUAV), designed to perform “over-the-hill” and “around-the-corner” reconnaissance
• Tactical UAV (TUAV), which focuses on the close battle, providing targeting, situation development and battle damage assessment in direct response to the brigade/Task Force commander

• Endurance UAV, aimed at the deep battle, supporting the division to 150 Km and the Corps battle to 300 Km. This class has two subclasses of the Endurance UAV:

• Medium Altitude Long Endurance (MALE) UAV, designed to operate at altitudes between 5000 and 25000 feet
• High Altitude Long Endurance (HALE) UAV, which are designed to function as Low Earth Orbit satellites.

From a logical standpoint, the subclasses of UAV are disjoint. A UAV cannot belong to more than one subclass. There exists a resource list and schedule of available platforms, which shows the following UAVs available:

• A MQ-2 Pioneer, which is a TUAV
• A MQ-1 Predator, which is a MALE-UAV
• A MQ-4 Global Hawk, which is a HALE-UAV.

Now suppose that as part of a given mission a Persistent Surveillance task over a wide area is required to detect any suspicious movement. This kind of tasks is best served by an Endurance-UAV, since it is able to fly for long periods of time. From just the concept definitions we know that:

1. the Pioneer is not an endurance UAV (because of the disjoint relationship among Endurance-UAV and TUAV)
2. both the Predator and the Global Hawk are Endurance-UAVs (because of the subclass relationships).

Both the Predator and Global Hawk meet the basic requirements. An additional rule checks the weather forecast, and determines that storms are likely during the planned mission time. That links to another rule, which states that in the event of bad weather, assuming the platform has a weather-penetrating sensor, a platform should be selected that can fly "above" the weather. In other words, a platform with high-altitude capability is needed. The Global Hawk is the only available platform that meets all these requirements.

To go to a finer-grained level of matching, the project used information containment relationships, with examples from the ISTAR domain. Even beyond that technique is ordinal ranking of matching.

"Q denotes a query which specifies some intelligence requirements to be met, and S1 − S5 denote the specification of ISR assets (sensors and sensor platforms) to be matched against Q.

"our query specifies two basic requirements to be met:

1. Provide Infrared (IR) Imagery
2. Carry out a Night Reconnaissance task"

Their article describes the rank ordering, with an exact match of S_n to Q, a perfect match of the requirement to the collection platform, down to the other entirely. A less desirable alternative meets the flight profile requirements, but it carries synthetic aperture radar rather than IR, and a platform that only has visual-spectrum television and no night capability is completely unsuited.

It is to be noted that the requirements are what are critical, not the particular platform. For the specific requirements, they also might be met with a manned long-endurance aircraft (e.g., P-3 Orion or Nimrod R), or relays of aircraft, or with satellites with appropriate orbits and sensors. These were not included in the ontology used for demonstration.

TPED approach to advanced geospatial intelligence

Geospatial intelligence marries pictures from imagery intelligence (IMINT) to additional information, ranging from maps and other geographic data, to multispectral multispectral MASINT, to the locations of electromagnetic sources through MASINT and SIGINT. The problem is that of creating persistent "TPED (i.e., tasking, processing, exploitation, and dissemination of data) over vast geographic areas and at the time intervals of interest." These were grouped into six classes of challenge:

1. achieving TPED
2. compressing the time line of geospatial intelligence generation
3. exploitation of all forms of intelligence
4. sharing geospatial information with coalition and allied forces
5. supporting homeland security
6. promoting horizontal integration among intelligence disciplines

The heart of these "hard challenges" is that the different sources and methods use different ontologies. If a generalized geospatial intelligence ontology (e.g., a concept dictionary, thesaurus, concept taxonomies) could be developed, it would greatly aid geospatial analysis.

Ontologies for Open Source Intelligence

A challenge for Open Source Intelligence (OSINT) is the immense number of potential sources available on the Internet, including the "deep web" of subscription or pay-per-use resources. If something is available on the Internet, unless it is an expensive part of the deep web, it is apt to be faster to access than print sources, but not everything in print is available on the Internet.

Informal ontological approaches have been used that map keywords and phrases in the intelligence requirement to potential search engines and other resources, such as archived mailing lists that may not be accessible to search engines. Unless there are time or logistical constraints that prevent visits to libraries, document archives, etc., the ontology also needs to indicate those sources.

Another kind of information that the ontology can use is pre- or post-processing for search engines. For example, the entry of the requirement could not only suggest search engines, but potential search arguments; the methodology might create arguments that reflect the strengths and weaknesses of various searching tools. It also might create scripts or use tools that automate the task of retrieving the sources found by the searches.

Most "user-friendly" search engines do not have more powerful information retrieval tools such as proximity search, set operations, etc. These may be available, however, on the analyst's computer or network, into which the multiple retrieved documents can be stored.