Data has emerged as a key asset in our contemporary society. Due to independently stored data being unable to fully realize its value, a well-established data trading scheme is required to incentivize stakeholders to share their data. However, existing works only focus on the direct data sellers, without considering that the data sold may be derived data through processing. This one-sided profit sharing model lacks fairness, neglecting the potential contributions of data processors. In this paper, we discuss the processing and profit sharing of derived data in the data marketplace for the first time and propose a secure multi-version data trading scheme called SMDT based on blockchain. More specifically, we introduce a new role of “processor” into data trading. Data owners are encouraged to provide raw data to processors, who then optimize the data into a derived version. When trading derived data, both the contributing processors and data owner can obtain revenue. To achieve fair and trustworthy profit sharing, we design a new smart contract called TMPS, enabling Traceable Multi-party Profit Sharing. Furthermore, our scheme ensures that data owners have absolute ownership over the multi-version data, while processors can only trade their derived data, without stealing the detailed content. Through experimental analysis, we validate the advantages of the proposed scheme in terms of security, usability, and efficiency.