Automated Detection Method for Privacy Policy Completeness of Mobile Applications

The privacy policy of mobile applications plays a crucial role in assessing whether the collection and use of personal information comply with regulatory requirements. Recent laws and regulations have raised the bar for how mobile applications should handle users’ personal data, yet existing research frameworks for evaluating privacy policy completeness fail to fully account for these updated legal and regulatory standards. Furthermore, issues such as unreasonable label design and insufficient coverage of relevant topics persist. To address these challenges, this paper proposes a new index system for evaluating the completeness of mobile application privacy policies. Building on this index framework, we create a dataset of privacy policy phrases and clauses, enabling the automated detection of policy completeness. Specifically, by analyzing the key requirements from six major regulations, we have developed an index with seven primary labels, 22 secondary labels, and 23 tertiary labels. For automated policy completeness detection, we constructed a dataset consisting of 10, 088 privacy policy paragraphs and 14, 606 clauses, training a deep learning model to classify these elements with accuracies of 88% and 90%, respectively. Using the completeness index and classification results, we evaluated the privacy policies of 400 mobile applications and found that 78.25% of them contained incomplete language segments, while 94% had missing or incomplete clauses.