Publisher's Synopsis
Today's systems rely on a collaboration of nodes (servers, mobile devices etc.), where selfish operators may violate system rules to gain personal benefit. Such behaviour has crucial impact on the system and is indeed considered as attack. Traditional techniques can weaken but not prevent the harmful impact as selfish nodes may hide their behaviour behind arbitrary faults. This work introduces an approach to reach a targeted level of collaboration for distributed systems being deployed over selfish nodes. A monitoring logic verifies a node's behaviour, permitted to apply a penalty in case of a detected misbehaviour. The logic copes with complex & dynamic situations where nodes are aware of the system state & other nodes' preferences and thus able to perform strategic choices consequently. This complex & interdependent setup corresponds to a class of game theory: Inspection Games. It is used to model the nodes' behaviour and determine security parameters correspondingly. The approach is designed as generic framework and can be applied to any architecture, selfish goal and reliability technique. A Publish/Subscribe system serves as illustrative use case for the detailed discussion.
