International Journal of High Performance Computing and Networking
| Hi-index | 0.00 |
Computational grids are characterized by the collaborative work among environment devices. Further, grid applications have been built based on reputation solutions. However, those applications can suffer due to attacks from malicious nodes. These nodes may not only corrupt the result from processed jobs, but also intentionally acquire a good reputation so as to obtain privileges to damage other nodes. In order to detect and isolate these malicious nodes (known as intelligent cheating nodes) from a P2P grid computing, this work proposes a system-level diagnosis model, using a strategy based on voting and honeypots. The model is evaluated by means of scenarios that take into account different percentages of malicious and cheating nodes. Achieved results show the robustness and efficiency of our diagnosis model, once all cheating nodes can be detected, with an accuracy of 99.4% of jobs being correctly processed. Furthermore, a graphical user interface is implemented for visualizing the simulations. Copyright © 2009 John Wiley & Sons, Ltd.