Planning Big-Data Distributed Processing for Combat ID

NPS NRP Executive SummaryProject Summary: Combat identification (Combat ID) of aircraft and ships is an important problem for militaries. Aircraft in particular move quickly, may not aid identification, may be subject to system malfunctions, or may be misconfigured. Identification is also becoming more difficult because commercial vehicles are less often in air lanes, and we now have autonomous aircraft in the air on nonstandard routes. We now have basic information about most major aircraft from satellite coverage, so combat identification today is a big-data problem. Even if we have facilities to handle large amounts of data, often we have a limited bandwidth to transmit sensor data to them. Our strategy is to push some of the processing and intelligence "to the edge" or to the platforms that collect data. We are developing methods whereby important data can be identified at an early stage. We have identified 16 factors which should be reported when they have anomalous values such as speeds, altitudes, counts in particular areas, deviations from tracks, and mixes of aircraft types in an area. Full data will still need to be transmitted eventually, but early forwarding of time-critical information could help neighboring platforms if we can identify it. We show some results from a prototype implementation using a sample of 110 million records from the ADS-B database. Results on a Hadoop distributed-processing system show speedups of 2-70 over a single-processor implementation depending on the subtask, which says this problem is well suited for distributed big-data methods.N2/N6 - Information WarfareNPS-18-N193-BApproved for public release; distribution is unlimited