In order to improve the protection of personal data in data lakes, we propose a meta-model that includes several aspects of privacy. Our meta-model describe all necessary constraints for the implementation of personal data protection procedures (pseudonymization/anonymization) in data lakes. This meta-model also enhances GDPR compliance in data lakes by having, for example, a personal data processing log and the identification of the finality of each data integration. Our meta-model is presented via a conceptual schema (UML) and implemented via a graph database (Neo4j). The validation of our proposal is illustrated by modeling and discussing several personal data protection scenarios in data lakes.