Users provide their own reputation value to a central server. The system can easily be cheated by providing a high reputation regardless of contribution (Kazaa Light client).


Point system where points are consumed by downloading and earned by uploading. A central server keeps track of every transfer and each user's total points. Uploading gives more points than downloading takes away. Peers can download at least 3 GB of data before download speed is regulated / downscaled.


Tit-for-Tat mechanism where a peer favors uploading to peers that provide the fastest rates in return. No long term history is kept.


Volume based tit-for-tat where a long term private history of transfers is kept. A peer favors uploading to peers that have provided the most data in the past.


Variant to eDonkey. A long term private history of transfers is kept at the eMule central servers. Peers favor uploading to peers with a high upload/download ratio. Credits are kept at the servers to prevent tampering with one's own credits. An exception can be given to friends, who can be assigned a "Friend Slot" regardless of credits.


A distributed reputation system, designed to counteract content pollution in peer-to-peer filesharing systems. Credence enables a peer to determine the authenticity of online content. Participants in the network vote on objects. Credence collates these votes, and weights them by a similarity measure which weighs highly votes from like-minded peers while discounting the votes from peers engaged in vote-spamming. This voter correlation scheme provides an incentive for peers to vote honestly and mitigates the impact of dishonest peers.


Distributed computation of globally consistent trust values. If M represents a matrix of direct trust valuations between pairs of peers, EigenTrust is a secure, distributed mechanism to compute M to the power n for all peers using power iteration. Distributed power iteration mechanisms are available.

Multi-level Tit-for-tat

A variation of EigenTrust where peers are compared first according to the interaction distance they have to the valuating peer (friend, friend-of-friend, etc.). Peers with the same distance d are ranked according to the matrix M to the power d. This reduces the effect of hitchhiking, and provides localized trust valuation in clusters that have little upload to the rest of the system.


A distributed economic framework where every peer has a globally consistent karma value. The system uses a DHT to keep track of objects and peers. Every peer has its own set of bank nodes which keeps track of its karma. When a peer desires an object the provider with the lowest price' is chosen; a payment' is made through the bank set before the object is transferred. All transactions use cryptographic mechanisms for security. A new peer is awarded a seed amount of karma to start with, and it automatically becomes a bank node for a selection of other peers. Periodically, bank sets perform corrections for inflation and deflation of karma.


Distributed accounting system where every node keeps a debit/credit account as well as a confidence value for a limited set of overlay neighbors. Newcomers receive a small amount of initial credit from some others. Scrivener relies on strong identities that cannot be whitewashed. If nodes don't have an overlay relationship but want to transfer data, a transitive trade has to be performed where credit is accounted on the path from the source to the destination.

Reputation (Gupta et.al.)

This `partially distributed' system proposes to use a reputation computation agent (RCA) to store peer reputations. Different reputation computation schemes based on credit and debit are suggested. As the RCA is necessary to prevent cheating, and a situation of multiple RCAs is not discussed, this solution is in essence a centralized solution.

One hop reputation protocol

This protocol exploits the small-world effect (just like BarterCast). Clients maintain a persistent history of interactions. They can also serve as intermediaries attesting to the behavior of others. Intermediaries serve two purposes: bootstrapping connections between new peer pairs and maintaining accounting information regarding indirect reciprocation. When two peers meet they first find out which peers they share in common in order to determine the relevant intermediaries. A number of (signed) messages are introduced to claim security.