The continuous production of large-scale data makes data outsourcing computation a trend. In order to ensure that data outsourcing transaction is trusted and fair, it needs the supervision and judgment of a third party. However, the third-party intermediary increases the transaction cost, and there is also subjective unreliability. Smart contracts allow trusted transactions without a third party, which are automatic, traceable and irreversible. Therefore, in this paper, we propose a framework for data outsourcing computation transaction based on the smart contract, in which the transaction is verifiable, fair and privacy protected. In this framework, by improved the replication-based verifiable computation technology, the data in the dataset is evenly distributed to each server, and at least one data in the sub datasets of the adjacent servers is the same, only the results from the duplicate data are verified, so as to implement the verifiable transaction with the low transaction cost. Meanwhile, a punishment mechanism is adopted to solve the fairness of the transaction in the smart contract. In addition, the oblivious transfer protocol is used to implement the privacy protection of transaction data. We deploy the data outsourcing computation transaction framework in the simulation environment of the Ethereum blockchain, and the experimental results show that our proposed scheme is effective and has low overhead. Specifically, the data returned by the server is verifiable, and the data privacy of the client is protected. When the client and servers are dishonest, they will be punished so that the transaction is fair. Moreover, the cost of using the smart contract in transactions is almost negligible.