Noblis, Inc, a leading provider of science, technology, and strategy services to the federal government, today announced the award of U.S. patent 11,423,157 for a system that simulates security checkpoint environments, using machine learning to model how different security configurations will respond to a variety of threats.
Using machine learning techniques, we’ve seen that computer systems can teach themselves rule-based adversarial games like chess, like chess, a security checkpoint is a battle of threat and response in which the outcome depends on the strategies the opponents choose. If an algorithm can teach itself chess, we wanted to see if we could use the same technology to learn interesting things about the adversarial environment of a security checkpoint.” Stephen Melsom, Noblis engineer and co-inventor of the system.
Instead of training on massive amounts of data, Noblis researchers give the system the rules of the game and ask it to play out the possible outcomes of an attacker trying to smuggle weapons or dangerous materials through a security checkpoint.
The system can identify capability gaps in each configuration of rules, Moreover, it can calculate the return on investment of various changes to that configuration so that decision-makers can use this technology as a strategic resource for allocating limited research, development, and operational funding.” Brian Lewis, Noblis team lead and co-inventor.
This invention came about through our Noblis Sponsored Research program, which we use to invest in the research and development of promising technologies, because we have both the domain and technical experience related to security operations in-house, we can quickly move from idea to solution.” Chris Barnett, Noblis’ chief technology officer.