Cyberattacks can lead to data breaches, service disruptions, and economic losses, and may even threaten national security and social stability. Therefore researchers have proposed various methods based on public datasets to improve the intelligence and automation of cybersecurity defense techniques. However, these public datasets usually have limited coverage of the types of cyberattacks, resulting in the proposed methods being ineffective against attacks not included in the dataset. Meanwhile, cybersecurity defenders often need to study cyberattack scenarios involving specific assets that are usually not represented in public datasets. To address these challenges, we propose a new approach to cybersecurity controlled dataset generation. Our method can reproduce any cyberattack using our four-role architecture, generating customized private attack data that includes specific assets, this capability satisfies the needs of researchers. By integrating the private attack data with a cybersecurity knowledge base derived from open-source datasets, we construct a comprehensive cybersecurity dataset. Extensive experiments demonstrate that the cybersecurity dataset generated by our method is suitable for various common cybersecurity tasks, such as threat hunting, alert analysis, and knowledge reasoning.