Industrial big data has experienced from data silos due to its high potential value and strong security requirements, making it difficult to share securely across domains. Blockchain-based solutions allow nodes to establish access control to trusted data on unreliable or trustless networks, but still face issues such as inefficient data sharing and leakage of sensitive information. In this paper, we propose a blockchain-based access control scheme for privacy security and dynamic regulation. First, ciphertext policy attribute-based encryption (CP-ABE) is developed to gain fine-grained access to node resources, with verifiable outsourcing decryption method to significantly reduce computational pressure on end users. Second, a policy hiding method based on multi-chain architecture is proposed, which performs double hiding of attribute information and access policy information on the blockchain. Finally, a supervisory policy that incorporates dynamic trust assessment and smart contracts is proposed to achieve effective detection and hierarchical classification punishment of malicious behavior. Security analysis and experimental results show that our scheme can limit the complexity of terminal decryption at a constant level and effectively achieve secure and efficient access control in the industrial Internet environment.