Case Study: Adversarial Networks

Needle in a haystack

Finding bad actors is a challenge. It is also crucial. Members of our team have worked with a defense contractor to prototype and implement AI-based systems to locate the proverbial needle.

Large datasets

The first engineering feat was to navigate the challenges of large data flows to achieve:

• Fully distributed Big Data infrastructure with decentralized persistence and predictive analytics
• Constant bidirectional synchronization protecting clearance levels at every node, secret, top secret and higher
• All local points down to laptops and mobile devices apply machine learning algorithms against the last synchronized device data set

Of large and small relationships

To identify obfuscated patterns calls for an analysis, our learning algorithms focused on micro-relationships within large and complex datasets. This resulted in multi-layered, multi-dimensional, indirect correlation and prediction algorithms that could probe and distinguish deep dependencies hidden in the data.

System goals

• To dynamically reduce complex events down to atomic correlations based on single attributes
• To perform automated algorithmic analysis of all relevant attributes taken together over time that define the progression of the set of interrelated events
• To conduct high-performance parallel processing — eliminating noise from the data stream to create the operational dataset
• To extract microscopic time-based pattern grids of the environment and applying adjustments as new input arrives in the data stream

Contact us to learn how we can customize AI for you.