Protótipo de apoio à conformidade com a LGPD: detecção automatizada de dados pessoais em portais acadêmicos

Abstract

The digital transformation of higher education institutions has expanded the availability of academic documents and informativo in online environments, but it has also increased the risk of improper exposure of personal data. In this context, the objective of this study was to develop and evaluate a prototype capable of identifying occurrences of CPF and RG in public documents from educational institutions, thereby supporting monitoring practices and compliance efforts related to the Brazilian General Data Protection Law (LGPD). To achieve this, an applied and experimental approach was adopted, integrating Web Crawling, Web Scraping, and Natural Language Processing techniques. A synthetic dataset was manually created and annotated to train a Portuguese-language NER model, which was later complemented by real-world tests based on search engine indexed results. In a controlled environment, the model achieved a precision of 0.967, recall of 1.000, and an F1-score of 0.983. In real-world settings, the solution proved useful as a document screening mechanism, reaching satisfactory average precision in identifying documents containing exposed personal data. The results indicate that the proposed approach is technically feasible and can support academic institutions in the preventive identification of improper data exposure, strengthening information governance and data protection practices.