Journal of Data Processing Wei Zeng 15 4 2025 https://doi.org/10.6025/jdp/2025/15/4/151-158 https://www.dline.info/jdp/fulltext/v15n4/jdpv15n4_1.pdf The paper proposes a cloud based data integrity verification algorithm for accounting informationization systems, leveraging rough set based data mining techniques. With the integration of cloud computing into financial management, data security and integrity have become critical concerns due to risks like data breaches and accidental loss as illustrated by real world incidents involving Huawei and Tencent Cloud. Traditional verification methods, such as HMAC or RSA based Provable Data Possession (PDP), suffer from high computational and communication overheads, especially under dynamic data operations. To address these limitations, the author introduces a parallel verification algorithm that supports multi user batch validation, significantly reducing communication costs and verification time. By employing rough set theory, the system performs efficient feature selection and attribute reduction, enhancing computational efficiency without compromising data integrity. The verification process uses BLS signatures with homomorphic properties, allowing multiple files to be verified at the communication cost of a single file. Experimental results show that the proposed algorithm outperforms conventional single point verification methods, especially as user scale and data volume increase. The system also offers tiered audit levels that balance detection accuracy, verification frequency, and cost. Implemented in a Java/SQL Server environment, the approach demonstrates improved scalability and suitability for accounting cloud environments, offering a practical solution for secure, efficient financial data management in the era of digital transformation.