news | 30 October 2024

Machine-Learning Powered Platform Provides DoD Ability To Identify Threat Network Activity

dMetrics_featured

$99.5M-ceiling award led by Defense Technical Information Center (DTIC), the U.S. Army, and Defense Innovation Unit provides greater access to dMetrics’s platform to support threat network detection.

Mountain View, CA (October 30, 2024) – U.S. adversaries leverage global systems of commerce and communications to expand their influence. Their activities are often traceable in the open source, but are difficult to capture, track, and analyze at scale by U.S. government analysts due in part to the vast amount of publicly and commercially available information (PAI/CAI) they must sift through, often manually, to accomplish their mission. 

To help analysts elevate critical signals from this increasing noise, DIU partnered with the Army and the Defense Technical Information Center (DTIC) to launch the AI-Based Knowledge Graphing project in 2019. The project aimed to develop a customized machine learning platform to rapidly ingest, analyze, visualize, and generate reports of strategic threat activity from open source, web-based content. A total of 65 companies submitted proposals in response to a DIU solicitation, and after a thorough evaluation process, dMetrics was selected in March 2020 to carry out a prototype project.

In just over a year, dMetrics successfully finished customizing their Minsky platform to ingest large amounts of data from diverse sources to reveal open source adversarial activity. dMetrics also successfully developed new capabilities, such as near-real-time alerts for specific behavior and the identification of early indicators of adversarial network activity, to support DoD analysts. The platform allows DoD analysts to create personalized machine learning (ML) agents that continuously scan big datasets to identify and extract entities, actions, and relationships relevant to each analyst’s area of responsibility. Having access to this capability has reduced the amount of time DoD analysts spend sifting through open source data by 90%. 

DTIC administrator Christopher E. Thomas said, "DTIC appreciates the efforts of DIU in identifying and helping dMetrics navigate the DoD’s procurement processes. The award of this production contract establishes a launch point for DTIC to employ Minsky to expand insight into DoD’s scientific and technical information, allowing our users to visualize connections among researchers, organizations, and topics previously unrecognized." 

Intelligence community (IC) agencies have also leveraged this platform to successfully combine large-scale, unstructured text datasets to enable geopolitical knowledge graphing, as well as to extract domain-specific insights from technical literature. 

As a result of the utility of this capability to U.S. government analysts, DTIC awarded dMetrics a five-year, $99.5M production contract in September 2024 to support the government in its efforts to make faster, better decisions based on large-scale, unstructured, publicly available data. This contract will ensure that DoD analysts continue to receive the support they need to identify adversarial activity online rapidly and at-scale. 

Moving forward, additional agencies will be able to make use of Minsky’s capabilities. Since Minsky encodes an analyst’s behavior regardless of their area of responsibility, dMetrics’s platform can be employed to support analysts throughout the federal government to analyze foreign news sources, map criminal network activity, or simply build a more powerful search tool. According to DIU Program Manager Jaime Fitzgibbon, “The dMetrics Minsky platform changes the AI tooling dynamic by putting subject matter experts (SMEs), rather than engineers, in the driving seat. In doing so, dMetrics empowers mission owners to respond to the ever-changing national security landscape quickly and effectively.”