Skip to main content

Table 1 Summary of transformations and privacy criteria used by variable type

From: Utilization of anonymization techniques to create an external control arm for clinical trial data

Variable types

Privacy Criteria

Transformations

(Anonymization method)

Identifiers

 

Suppression

Quasi identifiers

(k, ε)—anonymity

Noise (exponential mechanism)

Samplinga

Numerical variables

Differential privacy

Noise (Laplace mechanism)

Categorical variables

Differential privacy

Noise (exponential mechanism)

Metadata

(record order within a table)

Record order-based attacks

Shuffling

  1. aSampling was conducted as part of the exponential differential privacy mechanism and specific for the records that failed the k-criteria