In the era of artificial intelligence, data security and privacy protection have become a core focus. As a distributed ledger technology, blockchain faces the challenges of data silos and privacy leaks while working with massive amounts of data. To address these issues, this manuscript proposes a local differential privacy based on swarm learning (LDP-SL). The method firstly employs localized differential privacy to ensure that when data is collected and processed on the end node, personal private data is protected by introducing noise to prevent differential and inference attacks. Secondly, leveraging the characteristics of swarm learning, this manuscript achieves data sharing and the aggregation of collective wisdom without directly transmitting raw data, which effectively avoids the probability of data leakage. Furthermore, the manuscript designs a model combining convolutional neural networks (CNN), which can adapt to datasets of different types and dimensions. By introducing Laplace and Gaussian mechanism differential privacy techniques, it ensures privacy protection during the model training process. This method also does a lot of work on no independent and identically distributed datasets. Through experimental validation, the method proposed in this manuscript effectively utilizes massive data in blockchain for data analysis and decision support while protecting data privacy.