Opportunistic communications for emergency support

• Main Author: Gorbil, Gokce
• Other Authors: Gelenbe, Erol
• Published: Imperial College London 2013
• Subjects: 621.3
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.616843

In this thesis, we consider the problem of providing emergency support when existing communication infrastructure is unavailable. We propose using opportunistic communications (oppcomms) among mobile devices carried by civilians for the dissemination of emergency information. With oppcomms, devices exchange mes- sages at a close range of a few to tens of meters with limited or no infrastructure and messages are carried over multiple hops in a "store-carry-forward" manner by exploiting human mobility. We specifically look at the evacuation component of emergency response and propose an emergency support system (ESS) based on oppcomms to provide evacuation guidance to civilians in small-scale and large-scale urban emergencies in the absence of other means of communication. We evaluate the evacuation performance of ESS and investigate the communication characteristics of oppcomms for emergency support by simulation experiments. Our evaluations show that ESS improves evacuation by up to 31% and 14% compared to shortest path evacuation in large and small scale emergencies, respectively, and by up to 9% compared to a static-node based building evacuation system. We also investigate the resilience and security of oppcomms for emergency support under node failures and network attacks. We consider insider attacks where some nodes participating in oppcomms are compromised and misbehave. We investigate three different types of misbehaviour, including dropping packets, signal jamming and a hybrid attack on routing and evacuation that uses data falsification. Our evaluations show that node failures up to 20% are well-tolerated, and that data falsification has the most significant effect on evacuation by decreasing performance by up to 54%. In order to improve resilience of the system to such attacks, we propose a collaborative defense mechanism that combines identity-based cryptography and content-based message verification, and show that our defense mechanism improves performance by up to 50% in the presence of attacks.