Multidimensional social signature de-anonymizes low-sensitivity data.

Although privacy protection is increasingly important in the digital age, current measures predominantly focus on safeguarding high-sensitivity data, often overlooking the risks associated with low-sensitivity data. Here, we utilize low-sensitivity anonymous interaction data, which depict the behavioral interaction patterns between users and their contacts, to construct multidimensional social signature for identifying users in anonymous datasets. We investigate the potential of low-sensitivity social signature for user identification and propose a classification framework for measuring feature sensitivity levels. Among the test datasets, the accuracy of user identification can reach up to 87% in email communication, and remains wide applicability across different datasets. This suggests that even anonymous low-sensitivity interaction data should be considered personal data warranting protection under existing data protection regulations. We challenge the efficacy of current data anonymization methods and offer new perspectives on low-sensitivity data privacy protection through our feature sensitivity classification framework.