Network-aware Evaluation Environment for Reputation Systems
Parties of reputation systems rate each other and use ratings to compute reputation scores that drive their interactions. When deciding which reputation model to deploy in a network environment, it is important to find the most suitable model and to determine its right initial configuration. This calls for an engineering approach for describing, implementing and evaluating reputation systems while taking into account specific aspects of both the reputation systems and the networked environment where they will run. We present a software tool (NEVER) for network-aware evaluation of reputation systems and their rapid prototyping through experiments performed according to user-specified parameters. To demonstrate effectiveness of NEVER, we analyse reputation models based on the beta distribution and the maximum likelihood estimation.
Alessandro Celestini, Rocco De Nicola, and Francesco Tiezzi. Network-aware Evaluation Environment for Reputation Systems (Long Version). Technical Report, IMT, 2012.
The current version of the tool can be found here: NEVER.zip
- Java SE 6
Once decompressed the zip file ‘NEVER.zip’ open the new folder ‘NEVER’. Inside the folder there is a file named ‘Manager.jar’ and the following folders:
The folder ‘conf’ contains the configuration file ‘config.properties’ that is used to set evaluation parameters. The folder ‘data’ contains the file ‘User.jar’ that implements system’s parties. Finally, the folder ‘logs’ contains all the file returned as output by NEVER.
To run NEVER invoke the following command from the ‘NEVER’ folder
java -jar Manager.jar ExperimentManager
For suggestions, remarks, bugs or requests please do not hesitate to contact any of us.