Abstract:
Grid computing is widely regarded as a technology of immense potential in both industry and academia. Recently, the high computing industries like finance, life
sciences, energy, automobiles, rendering, etc. are showing a great amount of interest in the potential of connecting standalone and silo based clusters into a department and sometimes enterprise wide grid system. Even when the enterprises have considered grid as a solution, several issues have made them reconsider their decisions. Issues related to application engineering, manageability, data management, licensing, security, etc. have prevented them from implementing an enterprise-wide grid solution. As a technology, grid computing has potential beyond the high performance computing industries due to its inherent collaboration, autonomic, and utility based service behavior. To make this evolution possible all the above mentioned issues need to be solved. As an issue, security is perhaps the most important and needs close understanding as grid computing offers unique security challenges.
The work presented in this thesis is related to anonymous communication on grids. As grid computing scales up in size and diversity, anonymous communications will be desirable, and sometimes vital, for certain applications. However, existing anonymity protocols, when being applied to grid applications, either dramatically degrade system efficiency or cause severe performance bottlenecks. A distributed and highly efficient anonymity protocol can be designed if one considers the existing trust in grids. We have designed such a protocol based on controlled anonymity, which maintains an entity's anonymity against un-trustable entities. This protocol provides sender anonymity. We use the entropy based information theoretic based metrics to quantitatively analyze the degree of anonymity that could be offered by this protocol, and use a simulator to confirm its efficiency advantage over its predecessor, the 2-hop forwarding protocol
