Bridging the gap: Unlock the power of AI for government agencies through cross-domain solutions

Many organizations have already begun to realize the benefits of artificial intelligence, from automating routine tasks to generating new insights that accelerate decisionmaking. However, as the use of AI across government missions expands, the complexity of the digital environments they operate within presents unique challenges.

Government data is highly segmented by design, often separated by security classification levels to protect sensitive data and operations. While this segmentation is essential for national security, it also presents data-sharing obstacles that must be overcome. Fortunately, Cross-Domain Solutions (CDS) can help overcome obstacles such as safely training AI models with untrusted data, sharing classified AI capabilities with partners and connecting users or systems to AI tools across classification boundaries. 

Overcoming network isolation and data segmentation

Because government missions often operate across multiple classified and unclassified networks, users and systems may have access to data sources in an unclassified or protected environment, while a powerful AI tool is only accessible within a highly classified environment. This is due to some AI models being trained on proprietary or classified datasets or being capable of making decisions that reveal operational patterns. As a result, many AI tools are intentionally deployed only in highly controlled, classified environments. 

But there can be times when such AI models benefit from accessing data that is sourced from an untrusted environment, such as the internet. In these instances, agencies must have confidence that the AI model in a highly classified environment can safely access useful data across different classification boundaries without compromising the security of any apps, data and the AI itself. CDS can provide the secure bridge to enable these activities.

The Committee on National Security Systems has defined CDS as a form of controlled interface that provides the ability to manually and/or automatically access and/or transfer information between security domains. Originally developed to support information sharing between isolated classified networks, CDS have matured into advanced, security policy-enforcing technologies that safeguard against sophisticated and evolving cyber threats. Modern CDS serve as the secure interface between data sources, systems, users and powerful AI tools that are often separated by classification boundaries. 

Accelerating the OSINT lifecycle with CDS and AI

To understand how CDS and AI can work together to deliver real-world, immediate impact, consider a use case related to Open-Source Intelligence (OSINT) that illustrates how CDS can improve the analyst experience while securely enabling AI tools to accelerate the intelligence lifecycle.

Imagine an intelligence analyst working in a classified environment who receives a task to monitor social media chatter and commercial imagery for signs of activity in a particular region. The relevant data sources — such as news feeds, social platforms and mapping tools — are unclassified and publicly available, yet the powerful AI tool at the analyst’s disposal resides on a classified network. This example introduces two areas of friction that classification boundaries pose to the intelligence lifecycle. First, there is the analyst’s need to access unclassified resources while operating in a classified environment. Second is the ability to quickly feed data from the unclassified space into the classified space where it can be processed by the AI tool.

With CDS, this workflow becomes significantly more efficient. Certain CDS access solutions can allow the analyst to interact with unclassified sources from a classified workstation, in many cases eliminating the need to switch between workstations. And a CDS transfer mechanism automatically and securely transmits imagery, video, files or text into the classified environment for processing.

By incorporating CDS in the mission workflow, the cycle from collection to processing to analysis can be compressed from hours to minutes. Powerful AI tools can extract patterns, flag anomalies or prioritize items for review in a fraction of the time it would take a human. Combining these technologies within the intelligence workflow is essential to increasing the trust of data, improving situation awareness and creating a tangible decision advantage for stakeholders. 

A smarter, faster and more secure future

As the federal government increasingly turns to AI to accelerate operations, classification boundaries must be addressed thoughtfully and securely. CDS offers a security policy-enforcing method that is compliant with stringent government requirements to enable secure interoperability between AI and the data it needs. While continuing to assure isolation between classified boundaries is important, the alignment of CDS with data-centric security models and data modernization efforts will be a key factor in the evolution of CDS to support AI-driven workflows at scale now and into the future. 

For defense and intelligence communities in particular, the ability to safely operationalize AI across classified boundaries is a game-changing step toward smarter, faster and more adaptive tools that support mission success.

Christopher Finch is a solutions architect at Everfox. Dave Flanagan is a senior principal solutions architect at Everfox.