HRL Laboratories, LLC’s Complex Analytics of Network of Networks (CANON) system has achieved its second phase in the Modeling Adversarial Activity program (MAA), funded by the Defense Advanced Research Project Agency (DARPA).
A network we are analyzing may not be 100% correct due to noise, so we are anticipating noisy input records and designing the system to handle them accordingly
CANON is a set of software tools that can “do the math” to merge and analyze large graphs representing structured knowledge. Using integrated information from such networks of networks, CANON could be used to analyze and flag questionable adversarial activity.
“For the system to track down adversarial behavior we try to look data from a global perspective,” said principal investigator Jiejun Xu of HRL’s Information and Systems Sciences Laboratory. “Indicative activity that is not visible on the local level may be more obvious when viewed on a global scale, where data and information are properly fused together. This emphasizes the importance of generating a worldview network for integrated analysis, which is a key research area in the MAA program.”
When asked how CANON will be improved in the new phase, Xu said, “For Phase 2 we’re talking about analyzing large networks with much greater detail. Scalability will be important for this phase. The system will handle more data, so we’re improving the algorithms for smarter mathematical calculations. Also, we know that with these networks we can’t put everything into one machine, so we are using a distributed computing paradigm. For the system to work we have to adapt our method to scale horizontally across parallel computers.”
Another problem the team must solve is that more prodigious data also gets more “noisy.”
“We will be handling these much larger, noisier networks,” said Connie Ni, a computer scientist on the HRL team. “What we define as noise are such things as missing connections, data errors, and missing or partial data. It can also be data ambiguities, such as when two people have the same name. You may not be able to distinguish them, causing a data error we read as noise. Thus, a network we are analyzing may not be 100% correct due to noise, so we are anticipating noisy input records and designing the system to handle them accordingly.”
A third emphasis in MAA Phase 2 is the ability to conduct semantic reasoning on networks. This is defined as the ability to determine that two events described slightly differently are the same event. For example, “The US president landed in Paris in December” and “Trump arrived in France in December” are very likely the same event. The CANON system needs the ability to merge or align these two events in a worldview network.
Work in CANON would be carried out using a diverse set of synthetic networks to emulate real-world scenarios. The CANON team for the second phase of MAA also includes HRL scientists Tsai-Ching Lu, Alexei Kopylov, and Shane Roach as well as HRL’s institutional partner the University of Illinois at Urbana-Champaign. A workshop series on related graph analytic topics is co-organized by HRL annually. For more information, please visit https://gta3.ccni.hrl.com.
This material is based upon work supported by the United States Air Force and DARPA under contract number FA8750-17-C-0153. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the United States Air Force and DARPA.
HRL Laboratories, LLC, Malibu, California (hrl.com) is a corporate research-and-development laboratory owned by The Boeing Company and General Motors specializing in research into sensors and materials, information and systems sciences, applied electromagnetics, and microelectronics. HRL provides custom research and development and performs additional R&D contract services for its LLC member companies, the U.S. government, and other commercial companies.
Media Inquiries: media[at]hrl.com, (310) 317-5000