Data are collected and regarded as valuable assets in many business domains. Their owner would not want to disclose them to the public due to their potential value. Distributed knowledge discovery techniques have been proposed which assume the cooperation of data owners even though they might not behave in a trustworthy manner. When a party decides to quit the cooperation in the distributed knowledge discovery, the other parties cannot continue the discovery task and hence they get some disadvantage due to the party’s betrayal. This paper is concerned with data privacy-preserving distributed knowledge discovery which gives penalty to the party who quits the cooperation in the discovery process. It proposes a blockchain-based distributed machine learning method which does not disclose the participating parties’ data and gives the penalty to betraying parties. The proposed method makes the participating parties communicate with each other via the smart contract on the blockchain network. It uses a blockchain-based incentive system to establish trust among parties and to improve the quality of discovery knowledge. The proposed method has been implemented with a smart contract on the blockchain and tested for a benchmark data.

数据被收集并视为许多业务领域中的宝贵资产。他们的所有者由于其潜在价值，不希望向公众披露它们。已经提出了分布式知识发现技术，该技术假定了数据所有者的合作，即使他们可能行为不可靠。当一方决定退出分布式知识发现中的合作时，另一方无法继续进行发现任务，因此由于一方的背叛而导致他们处于不利地位。本文涉及保护数据隐私的分布式知识发现，这种发现会给在发现过程中退出合作的一方以惩罚。它提出了一种基于区块链的分布式机器学习方法，该方法不公开参与方的数据，并给背叛方以罚款。所提出的方法使参与各方通过区块链网络上的智能合约彼此通信。它使用基于区块链的激励系统在各方之间建立信任并提高发现知识的质量。所提出的方法已通过智能合约在区块链上实现，并经过了基准数据测试。