In traditional data trading process, data is sent directly to the demand side, leading to resale risks. To solve this problem, some researchers have proposed trading platforms that trade the right to use data, while data reamins local. However, due to the data invisibility, how to verify whether the data features meet the requirements of the Data requester without compromising the data privacy becomes a key issue. To address the issues above, this work propose a scheme for batch verification of data feature requirement fulfillment, where data holders organize and deposit data features using sparse feature Merkle trees (SFMT), and multiple data feature zero-knowledge proofs are aggregated using SnarkPack aggregation technology for data, thus enabling efficient verification of data feature zero-knowledge proofs on the chain. In addition, to avoid redundant verification of the same data features, this work propose a multi-way sparse Merkle tree (MSMT) to achieve efficient storage and verification of zero-knowledge proofs. Experimental results based on this scheme demonstrate that our proposed scheme significantly reduces the size of zero-knowledge proofs on the chain and decreases the gas consumption for verifying data requirement proofs on the chain, and MSMT also shows better performance in terms of insertion efficiency and generation and verification of Merkle paths, which improves the efficiency of verifying proofs of data feature requirement satisfaction on blockchain.