Study of privacy aware distributed detection in a heating and ventilation system

• Main Author: BARREIRO MANERO, ALBA
• Format: Others
• Language: English
• Published: KTH, Kommunikationsteori 2015
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-166843

The recent studies on sensor networks are causing a huge development of wire-less communications. These studies are motivated by the wide-range promisingapplications. However, this development faces multiple challenges. The privacyissue is one of these challenges. The study on the privacy-constrained sensornetworks is presented in previous works where the eavesdropping privacy risk istaken into account in the physical-layer distributed detection design. Nonethe-less, a rigorous study on the receiver operating characteristics (ROC) is missing.In this thesis, we will focus on a parallel distributed detection networkwith one fusion node that makes the final decision. We show how the fusionROC changes against different settings when considering the privacy constraints.First, a Bayesian detection scenario is considered. Second, some parameters aremodified for the purpose of studying the affect on the ROC. Finally, an otherscenario is analyzed where the eavesdropping is modeled a Neyman-Pearsondetection attach while the settings of the Bayesian distributed with detectionsystem is kept. === De senaste studierna om sensorn ̈atverk ̈ar f ̈oranledd av en snabb utveckling inomtr ̊adl ̈os kommunikation. Studierna motiveras av de lovande till ̈ampningar somfinns inom ett flertal omr ̊aden. Utvecklingen av sensorn ̈atverk st ̊ar dock inf ̈orett antal utmaningar. Den personliga integriteten ̈ar en utav dessa utmaningar.Studien om sekretessbegr ̈ansade sensorn ̈atverk finns presenterad i tidigare verkd ̈ar integritetsrisken vid avlyssnings tas i beaktning inom det f ̈ordelade fysiskalagrets detektionsdesign. Dock saknas en rigor ̈os under ̈okning p ̊a mottagarensdriftegenskaper (ROC).Denna avhandling kommer att fokusera p ̊a ett parallellt distribuerat detek-tionsn ̈atverk med en fusions nod som fattar det slutgiltiga beslutet. Vi visar hurfusions ROC ̈andras med olika inst ̈allningar n ̈ar man tar h ̈ansyn till begr ̈ansadsekretess. F ̈orst studeras ett Bayesianskt detekteringsscenario. Sedan kommervissa parametrar modifieras i syfte att studera dess inverkan p ̊a ROC. Slutligenanalyseras det andra scenariot d ̈ar avlyssningen modelleras som en Neyman-Pearson detektion medan inst ̈allningarna f ̈or det Bayesiansk distribuerade de-tekteringssystemet fortfarande h ̊aller.