Method for evaluating the risk of re-identification of anonymised data
The method delivers a degree of protection (txP 3 ) representative of the risk of re-identification of data in the case of a correspondence search attack including a deterministic search based on an external information source and a correspondence search based on a distance. The method comprises steps of E) consolidating a set of original individuals (EDO) and a set of anonymous individuals (IA); F) identifying, in the set of original individuals, individuals at risk (IOrs) via the deterministic correspondence search; G) evaluating a degree of failure of re-identification (txP 1 ) for the sets of original individuals and of anonymous individuals, on the basis of the correspondence search based on distance; H) computing the degree of protection as a function of a total number of individuals in the original dataset, of a number (RS) of individuals at risk identified in step B) and of the degree of failure of re-identification (txP 1 ).
1. A computer-implemented data processing method for evaluating a risk of re-identification of anonymized data, said method delivering a protection rate parameter (txP 3 ) representative of said risk of re-identification in case of a correspondence search attack including a deterministic search based on at least one external information source and a correspondence search based on a distance, said method comprising the steps of E) consolidating an original dataset (EDO) comprising a plurality of original individuals (IO) and an anonymized dataset (EDA) comprising a plurality of anonymous individuals (IA), said anonymous individuals (IA) being produced by a process of anonymizing said original individuals (IO); F) identifying, in said original dataset (EDO), original individuals at risk (IO rs ) as being original individuals (IO) having at least one noteworthy, or unique, value in at least one considered variable, or at least one combination of noteworthy, or unique, values in a set of considered variables, in a deterministic correspondence search and to which only one respective close anonymous individual (IA prs ) can be associated by said deterministic correspondence search; G) evaluating a re-identification failure rate parameter (txP 1 ) for said original datasets (EDO) and anonymized datasets (EDA), from said correspondence search based on a distance between each of said original individuals (IO) and one or more of the nearest of said anonymous individuals (IA) identified by a method called “k-NN” method; H) computing said protection rate parameter (txP 3 ) as a function of a total number (M) of original individuals (IO) in said original dataset (EDO), of a number (RS) of original individuals at risk (IO rs ) identified in step F) and of said re-identification failure rate parameter (txP 1 ) obtained in step G).
2. The method according to claim 1 , characterized in that, in step F), an anonymous individual (IA) is considered to be one of said nearest anonymous individuals (IA p , IA prs ) of one of said considered individuals at risk (IO rs ) when 1) said anonymous individual (IA) has a variable with the same modality as a considered variable of said original individual at risk (IO rs ) in said correspondence search in the case wherein said variable is a qualitative variable, or when 2) said anonymous individual has a value for said considered variable that is equal to a tolerance range close to the value of said same considered variable of said original individual at risk (IO rs ) in the case wherein said considered variable in said deterministic correspondence search is a continuous variable.
3. The method according to claim 1 , characterized in that step G) comprises the sub-steps of a) linking said original dataset (EDO) to said anonymized dataset (EDA); b) converting (PCA, MCA, FAMD) said original individuals (IO) and said anonymous individuals (IA) in a Euclidean space (A 1 , A 2 ), with said original individuals (IO) and anonymous individuals (IA) being represented by coordinates in said Euclidean space (A 1 , A 2 ); c) identifying, for each of said original individuals (IO), one or more of said nearest anonymous individuals (IA) based on said distance, using the “k-NN” method; and d) computing said re-identification failure rate parameter (txP 1 ) as being a percentage of cases where one of said nearest anonymous individuals (IA k ) identified in sub-step c) for one of said original individuals (IO i ) is not a valid anonymous individual (IA i ) corresponding to said original individual (IO i ).
4. The method according to claim 3 , characterized in that said distance is a Euclidean distance.
5. The method according to claim 3 , characterized in that the transformation of sub-step b) is carried out by a factor method (PCA, MCA, FAMD) and/or using an artificial neural network, called “autoencoder”.
6. The method according to claim 5 , characterized in that said factor method is a “Principal Component Analysis” (PCA) method when said individuals (IO, IA) comprise continuous type variables, a “Multiple Correspondence Analysis” (MCA) method when said individuals (IO, IA) comprise qualitative type variables, or a “Factor Analysis of Mixed Data” (FAMD) method when said individuals (IO, IA) comprise “continuous/qualitative” type variables.
7. A data anonymization computer system (SAD) including a data storage device (SD) storing program instructions (MET) for implementing the method according to claim 1 .
8. A computer program product including a medium in which program instructions (MET) are recorded that are readable by a processor for implementing the method according to claim 1 .