مهندسی کامپیوتر | تمرین ۳

Removing names from a dataset is a familiar step in protecting privacy, but a name is only one of the ways in which information can point to a particular person. Several ordinary attributes, unremarkable when considered separately, may become distinctive when they are combined or linked with records available elsewhere. Consequently, deciding whether a released table is sufficiently protected requires considering what a potential recipient could learn from it together with other information, not merely checking whether a column headed “name” has been deleted.

Suppose a research team replaces exact ages with age ranges before publishing a table of journeys, while retaining precise departure times and small departure locations. The broader age categories reduce one kind of detail, yet a journey may remain recognizable to someone who already knows when and where a particular traveler left. Making every field less precise could reduce that risk further, although it might also destroy distinctions needed to study transport at specific times and places.

The challenge is therefore to relate a proposed release to both a plausible use and a plausible route to identification. A protected research environment with limited access may support an analysis that would be inappropriate for an unrestricted public table containing the same detail. Privacy assessment must accordingly examine the combination of data, outside knowledge, and access conditions, rather than treating the removal of a single obvious identifier as a complete procedure. The assessment also has a temporal dimension, because information unavailable to a recipient at release may later become accessible through another record or service. A responsible account of protection describes the assumptions behind the release instead of presenting a permanently safe label detached from those assumptions.