Skip to main content
notice

Doctoral Thesis Defense: Mohamad Mehdi

May 12, 2015
|


Speaker: Mohamad Mehdi

Supervisors: Drs. J. Bentahar, N. Bouguila

Examining Committee: Drs. P. Grogono, W. Hamou-Lhadj, O. Ormandjieva,
N. Faci, H. Rivaz (Chair)

Title: Trust and Reputation Management in Web Services: A Probabilistic Approach

Date: Tuesday, May 12, 2015

Time: 10:00

Place: EV 1.162

ABSTRACT

The emergence of service-oriented architecture, the competitiveness of nowadays global markets, and the agility of business processes are the main drivers for a substantial shift in how business is conducted. We can notice this shift by examining the growth of the number of business applications that are designed and deployed as autonomous and interoperable agent-based web services. Consequently, the web is currently stocked with web service providers that offer similar business functionalities. Additionally, a large number of service providers behave in a dishonest and often malicious fashion for either greedy or competitiveness reasons. These facts lead to a major problem that service consumers are presently facing; selecting the most appropriate and trustworthy providers. Therefore, to solve what is referred to as the web service selection problem, we explore and analyze the behavior of service providers to build a two-fold capability; (1) maximizing the gain of service consumers by selecting the service provider that best meet their quality requirements and (2) avoiding interactions with malicious providers. We base this capability on modelling and learning the quality of service (QoS) attributes of the service providers using probabilistic machine learning and data mining approaches. Specifically, we propose several QoS-based models to estimate the trustworthiness and aggregate the reputation of service providers. To render the solution practical and useful in real world applications, we extend our models to function in online settings. Finally, we evaluate these models by running various simulations that yielded very promising results in comparison with the state-of-the-art trust and reputation models.




Back to top

© Concordia University