Data fusion without data fusion: localization and tracking without sharing sensitive information

Abstract

It is now well known that data-fusion can improve detection and localisation of targets. However, traditional data fusion requires the sharing of detailed data from multiple sources. In some cases, the various sources may not be willing to share such detailed information. For instance, current allies may be willing to share some level of information, but only if they can do so without revealing their secrets. We show here that, at least for localisation and tracking of targets, data-fusion can be performed without the need to actually combine the data in question, so that no party learns the information of another. Such an approach would allow co-operation between parties that share mutual interests, and yet do not completely trust each other. The particular application on which we concentrate is localisation and tracking of a target using multiple sensors (radars or sonars, or other devices). We show that multiple sensors can be used to refine a target's position estimate, or even obtain a position estimate where no single sensor has enough information to do so (e.g., each has only range information), without sharing the details of the sensor, such as its position, or accuracy of its estimates. © 2007 IEEE.