Portable Reputations with EgoSphere Keith Bonawitz Chaitra Chandrasekhar Rui Viana bonawitz mit edu chaitra mit edu ruilov mit edu worth the time to read ABSTRACT Many online services require some form of trust between users trust that a seller will deliver goods as advertised trust that an author s thoughts are worth the time spent on reading them To accommodate an internet community where users are constantly interacting with strangers online services often construct proprietary reputation management systems for their community with the side effect of locking users into that service if they wish to maintain their reputation In contrast this paper outlines EgoSphere a system for portable Internet reputations so that reputations built on one service can be used elsewhere EgoSphere hinges on the use of correlational statistics to automatically project reputations from one service to other similar services To achieve these goals EgoSphere must gather webservices reputation data EgoSphere avoids the unreasonable expectation that all webservices will publish their reputation databases while also avoiding the use of a webcrawling robot a violation of many webservices robots txt restrictions and source of incurring additional website load by gathering its reputation data using a distributed passive robot system This system simulates the function of a standard webcrawling robot by using webproxies on users computers to analyze the responses to standard webservice requests This paper outlines the design and proof of concept implementation of EgoSphere targeted specifically at providing portable reputations for bulletin board style internet services To facilitate trust among strangers reputation systems have been successfully applied in various settings on the Internet These systems provide summaries of users pasts with the rationale that a user who has acted trustworthily in the past is likely to continue to do so Such summaries must be concise if a user s reputation summary is too lengthy reading it will cost more than the informational gain it provides In many cases these concise summaries take the form of a single number with the benefit that software can also use these numerical summaries to automatically organize and filter information The most notable reputation system currently deployed is Google s PageRank algorithm for computing the expected relevance of a hit from a web search e g how much the user should trust the hit to provide useful information High PageRank indicates high expected relevance and results from the page having a large number of other pages linking to it The implicit assumption is that links to a webpage are evidence that someone finds that webpage relevant Amazon zShops and eBay two of the most successful online marketplaces deploy typical reputation systems for sites based on commercial transactions When users complete transactions the system allows the users to rate each other how smoothly the transaction was executed Most of this information is made public so that anyone can easily understand the past performance of a specific user In order to demonstrate the relevance of reputation systems some research has already been made DEL01 on how e commerce reputation services can affect the online markets for the goods offered through these web sites Intuitively if a user holds a higher reputation rank he or she is able to sell products for a higher price 1 INTRODUCTION Reputation Systems With over 900 million people COM interacting on the web Internet users are regularly finding themselves in situations where they must choose to trust strangers trust them to faithfully complete a commercial transaction or to provide information that is 1 Other services such as Slashdot or infoAnarchy use reputations to assign trust to certain pieces of information At Slashdot users can rate the comments posted by other users Although Slashdot only publishes percomment rating summaries rather than peruser rating summaries it internally uses peruser summaries called Karma in Slashdot terms to make highly reputed users comments more visible by synthetically boosting their per comment reputation scores This in turn results in earlier placement of the comment on the web page Although not published a user s Karma can be inferred by reviewing a history of that user s percomment ratings EgoSphere proposes to integrate different reputations services into one global system facilitating the transfer of reputations between services Since EgoSphere gathers data from feedback provided from many different websites it has more information available to its algorithms than any single service does Thus EgoSphere can compute reputation rankings that are more informed than the ones computed by any single service Naturally the values provided by each service are still available to the user but the ability to gather information from other websites allows EgoSphere to provide a more complete reputation profile EgoSphere particularly shines in its ability to fill in reputations when users are new to a service until the user can build up a local reputation The current design of EgoSphere focuses on information sharing websites like Slashdot It is important to make clear what an overall reputation value would reflect on such sites If a user receives good feedback for one of his comments on the Apache section of Slashdot it could mean that the user is a skilled writer on technical issues or s he has a great knowledge about the Apache server It is our desire that the reputation value reported by EgoSphere will account for both of these components Isolated reputation systems leave no option to the user other than to perform most of his or her transactions at the dominant websites A good reputation built at a smaller online service is not nearly as valuable as one built at the larger websites Hence it is extremely difficult for smaller or new services to compete against the larger ones However if the reputations of a small service constantly agree with those of a larger service then there seems to be no reason why the users of the larger website should not trust the reputation values from the smaller as indicators of trust The Portable Reputation Vision A critical shortcoming of current reputation services is that they are generally bound to a specific website A user that has built a good reputation at Slashdot is not able to take advantage of that reputation at infoAnarchy even though these two websites