Abstract:
Most network applications such as pervasive computing, grid computing, P2P etc. can be viewed as
multi-agent systems which are open, anonymous and dynamic in nature. Due to such nature multiagent
systems present potential threats in ensuring secured communication. One of the key feasible
ways to minimize threats is to evaluate the trust and reputation of the interacting agents. Many trust
models have done so, but they fail to properly evaluate trust when malicious agents behave in an
unpredictable way. Besides that they are also ineffective in providing quick response to a malicious
agent’s oscillating behavior, i.e., they are unable to assess the true nature of an opportunistic agent.
To cope with the strategically altering behavior this thesis presents a dynamic trust model called
SECTrust where we analyze the different factors related to the trust of an agent and then propose a
comprehensive model for evaluating trust. In our trust model we have considered various factors in
computing the trust of agents which includes recent trend, historical trend, sudden deviation of trust,
confidence factor and decay of trust. We have also tuned our trust model for the different parameters
that we have considered to get the best possible results. Simulations show that our model compared to
other existing models can effectively cope with the strategic behavioral change of an agent and at the
same time efficiently isolate the malicious ones. So, our trust model is more robust and effective in
preventing attacks from opportunistic malicious agents and can thus ensure secured communication
among the participating agents.
