Access Type

Open Access Dissertation

Date of Award


Degree Type


Degree Name



Computer Science

First Advisor

Daniel Grosu


Most distributed systems comprise autonomous entities interacting with each other to achieve their objectives. These entities behave selfishly when making decisions. This behavior may result in strategical manipulation of the protocols thus jeopardizing the system wide goals. Micro-economics and game theory provides suitable tools to model such interactions. We use game theory to model and study three specific problems in distributed systems. We study the problem of sharing the cost of multicast transmissions and develop mechanisms to prevent cheating in such settings. We study the problem of antisocial behavior in a scheduling mechanism based on the second price sealed bid auction. We also build models using extensive form games to analyze the interactions of the attackers and the defender in a security game involving honeypots.

Multicast cost sharing is an important problem and very few distributed strategyproof mechanisms exist to calculate the costs shares of the users. These mechanisms are susceptible to manipulation by rational nodes. We propose a faithful mechanism which uses digital signatures and auditing to catch and punish the cheating nodes. Such mechanism will incur some overhead. We deployed the proposed and existing mechanisms on planet-lab to experimentally analyze the overhead and other relevant economic properties of the proposed and existing mechanisms.

In a second price sealed bid auction, even though the bids are sealed, an agent can infer the private values of the winning bidders, if the auction is repeated for related items. We study this problem from the perspective of a scheduling mechanism and develop an antisocial strategy which can be used by an agent to inflict losses on the other agents.

In a security system attackers and defender(s) interact with each other. Examples of such systems are the honeynets which are used to map the activities of the attackers to gain valuable insight about their behavior. The attackers want to evade the honeypots while the defenders want them to attack the honeypots. These interesting interactions form the basis of our research where we develop a model used to analyze the interactions of an attacker and a honeynet system.