Is your personal data safer to disclose? An exploratory analysis of reidentification risk

The ubiquity of personal data generated through human-centric devices—such as smartphones and wearable technologies—has intensified concerns over individual privacy and reidentification risk. Despite the implementation of data protection regulations that mandate strict disclosure controls, numerous studies have demonstrated the persistent vulnerability of de-identified datasets. In this study, I conduct a comprehensive risk assessment on a publicly available de-identified dataset, focusing on two dimensions of uniqueness-based risk: sample uniqueness and population uniqueness. The analysis reveals that, under an adversarial knowledge scenario, the probability of correctly reidentifying an individual record reaches 0.35. Furthermore, over 45% of records are susceptible to reidentification when seven quasi-identifiers are known, while even four attributes suffice to reidentify more than 9% of records. The proposed estimation framework achieves accuracy exceeding 75%, outperforming several baseline models. These findings highlight the limitations of existing anonymization techniques and underscore the need for more robust disclosure control mechanisms, particularly for datasets that involve sensitive personal attributes.