The matrix actually used for job seeker Some precautions data is more complex and guarantees that no disturbed data is between 1 and 4. The transition matrices used by INSEE in the context of the dissemination of QPVs in 2025 still guarantee that no disturbed value will be lower than the confidentiality threshold of the source (5 for job seeker data? 10 for health Some precautions insurance beneficiary data? etc.). Although from a theoretical point of view this property is not in itself absolutely essential to guarantee a satisfactory level of protection? it nevertheless contributes to it and also makes the protection provided by data disruption more tangible.
This method nevertheless calls for
Some new precautions in the use of data:
the maximum deviation being fixe? the small norway phone number library values are relatively more disturbed than the large ones: it is therefore preferable not to draw too strong conclusions from analyses relating to tables in which the proportion of these small boxes would be significant;
the decompositions of a total according to the underlying components are no longer exact. For example? in the file of job seekers? after perturbation the total number of job seekers does not generally coincide with the sum of the number of job seekers broken down by age group (whereas this is the case in the undisturbed data). This “loss of additivity” plays a role in this concentration of value at the mechanism of protection of secrecy: if a disruptive method of protection of secrecy sought to guarantee perfect consistency between a total and its decompositions? then the protection of certain decompositions (of small numbers in particular) would be weakene.
The dissemination of data protected with
The random key method is done aero leads with complete confidence:
it protects against both the risk of breach of confidentiality and suppressive methods while substantially reducing the risks associated with geographical differentiation:
