Title: Analysing indirect reciprocity in overlay networks

Abstract:

Overlay networks are built by installing standalone software in many
internet-attached hosts, thus creating a new network between these hosts
that uses the Internet as a substrate. With overlay networks it is possible
to have much greater control and flexibility on the creation and
maintenance of the network topology. However, this comes at a great
decentralization price: each node in the network is under the control of a
different administrative entity, and thus strategic behavior can become a
hindrance to the deployment of overlay network protocols. As a consequence,
it becomes necessary to design incentive mechanisms that allow strategic
peers to coordinate while still maximizing their own utilities.

A usual policy to achieve this is "Tit-for-Tat", where each peer
reciprocates to other peers that have altruistically provided services in
the past. Tit-for-Tat, however, does not scale well to dynamic networks
where peers interact basically among strangers. A possible technique that
can be applied in this case is indirect reciprocity, where the contribution
that a peer gets from the peer-to-peer system is provided by peers
different to those to whom it contributed in the past.

Although indirect reciprocity  is a well known technique in peer-to-peer
protocol design. However, the modeling and game theoretical analysis of the
contribution flows that make it possible are not well understood.
We propose techniques to use indirect reciprocity as an incentive mechanism
for peer-to-peer networks, and for the modeling and game theoretical
analysis of the contribution flows that make it possible. First, we present
an analytic framework for peer-to-peer contribution flow geometry and its
application to indirect reciprocity. This framework is based on the
adaptation of well-known differential operators to a graph-theoretical
setting. Second, we propose a utility function for QoS-aware peer-to-peer
overlays operating under indirect reciprocity, and calculate the peer best
response peer contributions as a function of the service quality that other
peers can provide and overlay topology parameters.