back to dictionary
K-anonymity
k-anonymity is a privacy protection model designed to prevent an individual from being identified from a dataset. To achieve this, it establishes that each record must be indistinguishable from at least k?1 other records with respect to certain attributes that could reveal a person's identity.
This concept was formalized by researcher Latanya Sweeney in 2002, in her scientific paper k-Anonymity: A Model for Protecting Privacy, published in the International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems. Her work addressed a fundamental data privacy issue: removing direct identifiers, such as names or identification numbers, does not always prevent a person from being recognized when combined with other attributes, such as age, gender, or postal code. The proposal built upon previous research on information disclosure and statistical data protection.
In Human Resources, this principle is particularly relevant in work environment surveys, where combining variables such as department, job position, or seniority can facilitate the indirect identification of participants, even when names are not collected.
A practical way to apply a similar criterion is to establish a minimum number of responses required to display a group's results. For example, if a threshold of five participants is set, results for teams that do not reach that number will remain hidden. This reduces the risk of re-identification when reviewing reports.
However, this threshold alone does not constitute a formal guarantee of k-anonymity. It is also necessary to consider potential filter combinations, available information, and the risk of reconstructing data through cross-result comparisons. Furthermore, k-anonymity does not eliminate all privacy risks, so it should be complemented with other protective measures.
The importance of these measures is also reflected in institutional practice. In a study on survey design and administration to improve organizational climate, published in Academic Medicine in 2022, researchers affiliated with the National Institutes of Health (NIH) described the use of aggregated results and the suppression of cells with fewer than 15 responses to reduce the risk of identification. This figure was established based on organizational characteristics, group sizes, and data sensitivity, rather than as a universal threshold.
Therefore, applying protection criteria based on k-anonymity helps reconcile two organizational needs: gathering actionable insights for decision-making and safeguarding employee privacy. In workplace environment assessments, this means designing reports and their filters so that analyzing results does not compromise participant confidentiality.
Sources and references:
