PhD Defense: Wei Dong, GDC 6.516

Contact Name: 
Lydia Griffith
Date: 
Aug 15, 2013 9:00am - 11:00am

PhD Defense: Wei Dong

Date: AUGUST 15, 2013
Time: 9AM
Location: GDC 6.516

Title: Collaborative Mobile Services

Advisor: Yin Zhang and Lili Qiu

Abstract:
Mobile devices (e.g. smartphones and tablets) are being adopted with unprecedented speed. The growth in device capability, demand and system complexity increasingly require collaboration of multiple parties in order to achieve new or better functionality, efficiency, performance, etc. This poses unique challenges such as information sharing among different parties, utility sharing among different parties, and dishonest and collusive behaviors.

In this work we first study the problem of information sharing. We consider friend discovery in mobile social networks, where users of a mobile social network work together with each other to discover potential new friends nearby by computing their social proximity. We identify attacks that compromise the privacy of users’ personal information and location via analysis of real social networks. We develop the first secure dot product protocols that simultaneously achieve privacy and verifiability. Our prototype implementation shows our approaches are efficient in both computation and power consumption.

We then propose auction based approaches for utility sharing in cellular offloading. We propose the first reverse auction framework - iDEAL that allows a cellular service provider to purchase different types of wireless resources from third party resource owners. iDEAL accounts for the diverse spatial coverage of different resources and effectively fosters competition. We also propose a double auction framework - DA^2 that enables dynamic spectrum allocation among cellular service providers. DA^2 takes a divide-and-conquer strategy and better captures the complex competition patterns (characterized by a conflict graph) among spectrum buyers. We theoretically prove that iDEAL and DA^2 are economically robust. Our trace-driven simulations with data from a major US ISP show that our auctions efficiently allocate resources from multiple parties and effectively incentivize participants, while mitigating dishonest and collusive behaviors.