Multiple Simultaneous Threats Detection in Distributed Systems

Abstract

This research examines a simultaneous threats detection system for distributed systems that uses a hybrid identification fusion model. This hybrid model is comprised of mathematical and statistical data fusion engines: Dempster-Shafer, Extended Dempster-Shafer, and Generalised Evidential Processing (GEP). The simultaneous threats detection system produced threat detection rates of 56% using Dempster-Shafer whilst Extended Dempster-Shafer and Generalised Evidential Processing (GEP) achieved 80% and 95% threat detection rate. Thus, the simultaneous threats detection system can improve threat detection rates by 39% (i.e. 95% - 56%) simply by adopting a more effective hybrid fusion model. In terms of efficiency and performance, the comparison of the three inference engines of the simultaneous threats detection system showed that Generalised Evidential Processing is a better data fusion model than Dempster-Shafer or Extended Dempster-Shafer. In addition, the set cover packing technique was used as a middle-tier data fusion tool to determine the reduced size groups of the threat data. Set cover provided significant improvement and reduced the threat population from 2,272 to 295. This helped to minimise the complexity of evidential processing, and therefore reduced the cost and time taken to determine the combined probability mass of the multiple simultaneous threats detection system. This technique is particularly relevant to online and internet-dependent applications, including portals.