Understanding the Democratic People’s Republic of Korea (DPRK), its capabilities, plans and intentions, has always been a particularly difficult challenge for outside observers, including Western intelligence agencies. Access to the country is limited. Information is tightly controlled. The regime has become extremely adept at countering or denying technical means of intelligence collection. According to Daniel Coats, the Director of National Intelligence, “It is one of the hardest, if not the hardest, collection nation that we have to collect against.”
This problem is even more acute with respect to the regime’s leader. Recently, The Atlantic published an article titled “Kim Jong Un: The Hardest Intelligence Target.” The author asks several questions that must be at the heart of the Trump Administration’s deliberations over how to respond to North Korea’s nuclear weapons program: “Is Kim Jong Un crazy or hyper-rational? Is he bent on destroying America or deterring America?” The answer is the only one any honest observer of Kim can provide: “Nobody knows.”
The DPRK is but one of a series of so-called hard intelligence targets. Russia, China, Iran and most terrorist groups are also considered hard targets. They are hard in the sense that customary collection methods often are not adequate to provide the desired quantity and quality of information. But even if access is available, it is often a challenge to correctly interpret the data collected.
Quantitative modeling of open source “big data” has the potential to widen the aperture with respect to hard intelligence targets. It could be particularly useful in gaining insights on how decision makers think and their internal operating dynamics. Quantitative models can identify policy trends early, help our understanding of decision making in secretive, closed governments, and support the development of strategies for dealing with target countries or leaders.
The most basic source of quantitative data is open source news stories. In the United States, print outlets such as The New York Times and The Washington Post produce between 250 and 500 news items a day. Television, cable and radio outlets each produce hundreds more. Academics and technical literature provide additional sources of news items that can be exploited, depending on the research question. In countries that impose press controls – which includes all the hard targets – fewer news stories are produced but their significance is greater because they are government approved.
Every day, just under 1,000 distinct stories are published on North Korea by a few hundred unique sources both inside that country and internationally. Each story is a source from which information can be extracted. Using models based on elite dynamics, which comb through the dataset to create statistical products based on thousands of observations, it is possible to penetrate the information barriers set up by this hard target.
An example is the purge of Jang Song-thaek, Kim Jong Un’s uncle and a powerful figure in North Korea for many years. Eventually, Jang’s fall from grace became public, but evidence of his fate was clear much earlier from the sudden lack of news stories about such an important individual. Jang’s absence was a fact observable based on the aggregation of many news stories but not apparent in any single article.
Open source data can be exploited using analytic models to study complex interactions both within and between nations. Elites operate by gathering and processing information and then selectively signaling to other elites and the public through their words and actions. How elites allocate their time—a scarce resource—is a potential indicator of future actions. The interaction between elites of two countries reported in open source news items can provide insights regarding attitudes and motivations, national strategies, the capacity of one elite to influence the other and even potential outcomes of bilateral and regional policy issues.
A pioneering company in the employment of dynamic statistical modeling of open source big data for intelligence purposes is Kingfisher Systems, Inc. The company created an innovative computation platform that combines machine-coded aggregate data with refined issue analysis to ascertain trends, and identify entities of interest. The platform has three major elements. First is the continuous autonomous collection, ingestion and storage of open source data from around the world. Some 200,000 news items are collected daily and stored for later exploitation. Second is a proprietary approach to the automated organization, integration and annotation of this large database. Third is the application of a unique computational analytic methodology, called VARYSS, to a specific intelligence or policy issue.
VARYSS exploits diverse and abundant sources of current and historical open source information to identify leading indicators related to significant events, strategic activities and historical trends. The collection of open source news items, technical information and other literature spanning over a decade allows a better understanding of events including complex and dynamic systems, on a global, regional, and national level. The collected data is validated, generally through automated processes.
Exploratory data analysis may be performed using descriptive statistics to support algorithm building and modeling using mathematical formulas. The algorithms or models enable the identification of relationships among variables often inaccessible by traditional analytic techniques. This often leads to more detailed research frequently employing statistical and mathematical analysis based on proprietary computational models. Data and completed analyses can be presented in a variety of visual formats.
Kingfisher Systems, Inc. employed its computational methodology to the study of North Korean decision making. Data and statistical models generated from a regional study of bilateral interactions were used for a more detailed look at DPRK’s political processes. What emerged was a perspective of North Korea as an oligarchy controlling a weak party that governs through a primitive patron-client system. As such, patrons extract loyalty by balancing distribution of meager resources with the use of political violence and intimidation. This perspective suggests the potential for alternative operational strategies by the U.S. to engage the North Korean regime.
Hard intelligence targets produce a surprisingly large amount of potentially useful open source information. Properly collected, organized and analyzed, this information not only can complement traditional intelligence collection methods but, in some cases, provide leading indicators. Quantitative modeling of open source data can improve U.S. intelligence analysis. It also can support the exploration of alternative U.S. policy options as well as test Pyongyang’s response to prospective initiatives.
Find Archived Articles: