Trustful ad hoc cross-organizational data exchanges based on the Hyperledger Fabric framework

Organizations share data in a cross-organizational context when they have the goal to derive additional knowledge by aggregating different data sources. The collaborations considered in this article are short-lived and ad hoc, that is, they should be set up in a few minutes at most (e.g., in emergency scenarios). The data sources are located in different domains and are not publicly accessible. When a collaboration is finished, it is however unclear which exchanges happened. This could lead to possible disputes when dishonest organizations are present. The receipt of requests/responses could be falsely denied or their content could be point of discussion. In order to prevent such disputes afterwards, a logging mechanism is needed which generates a replicated irrefutable proof of which exchanges have happened during a single collaboration. Distributed database solutions can be taken from third parties to store the generated logs, but it can be difficult to find a party which is trusted by all participating organizations. Permissioned blockchains provide a solution for this as each organization can act as a consensus participant. Although the consensus mechanism of the permissioned blockchain Hyperledger Fabric (versions 1.0-1.4) is not fully decentralized, which clashes with the fundamental principle of blockchain, the framework is used in this article as an enabler to set up a distributed database, and a proposal for a logging mechanism is presented which does not require the third party to be fully trusted. A proof of concept is implemented which can be used to experiment with different data exchange setups. It makes use of generic web APIs and behaves according to a Markov chain in order to create a fully automated data exchange scenario where the participants explore their APIs dynamically. The resulting mechanism allows a data-delivering organization to detect missing logs and to take action, for example, (temporarily) suspend collaboration. Furthermore, each organization is incentivized to follow the steps of the logging mechanism as it may lose access to data of others, otherwise. The created proof of concept is scaled to 10 organizations, which autonomously exchange different data types for 10 min, and evaluation results are presented accordingly