In edwindj/sdcSpatial: Statistical Disclosure Control for Spatial Data

knitr::opts_chunk$set(
   echo = TRUE,
   out.height = "70%",
   warning = FALSE
)
library(sdcSpatial)

sdcSpatial: Privacy protected maps

knitr::include_graphics("fig/privacy.jpg")

sdcSpatial: take home message

Experiments with sdcSpatial

• sdcSpatial [@dejonge2022, @WolfJonge2018],
• AT, DE, FR, NL experiments population density
• different utility measures tested
• different types of focus areas

sdcSpatial has methods for:

• Creating a raster map: sdc_raster for pop density, value density and mean density, using the excellent raster [@Hijmans2019].
• Finding out which locations are sensitive: plot_sensitive, is_sensitive.
• Adjusting raster map for protecting data: protect_smooth, protect_quadtree.
• Removing sensitive locations: remove_sensitive

Why sdcSpatial?

• ESS has European Code of Statistical Practice (predates GDPR, European law on Data Protection):
  no individual information may be revealed.

Sdc in sdcSpatial?

SDC = "Statistical Disclosure Control"

Collection of statistical methods to:

• Check if data is safe to be published
• Protect data by slightly altering (aggregated) data
• adding noise
• shifting mass
• Most SDC methods operate on records.
• sdcSpatial works upon locations.

What is sdcSpatial good for?

Protecting

• Spatial Population density
• Spatial value density, e.g. unemployment, income
• Spatial fractions, e.g. unemployment rate
• Spatial averages, e.g. average income

We'll focus on population density

Experiments with sdcSpatial

Using sdcSpatial for AT, FR, DE and NL to protect population density, i.e. to "grid locations" with < 10 persons [@Gussenbauer2023]

• 4 different area's per country: urban, moderately urban, rural, border area
• different utility measures, Hellinger distance, Moran's I
• 3 different methods: (next slides)

Example: unprotected

\scriptsize

population <- sdc_raster( dwellings[c("x","y")]
                        , variable = 1
                        , min_count = 10
                        , r = 500)
plot(population, value="count")

Cell removal: remove_sensitive

• Just remove the unsafe cells
• Pro: simple, fast, no artifacts introduced
• Con: looses mass, low density areas are removed

pop_removed <- population |> remove_sensitive()
plot(pop_removed, value = "count")

Smoothing: protect_smooth

• Spatial smoothing, unsafe locations are "blurred"
• Pro: highlights spatial pattern, removes spatial noise
• Con: Loose mass at edges, Introduces unplausible locations.

pop_smoothed <- protect_smooth(population, bw = 500)
plot(pop_smoothed, "count")

Quad tree: protect_quadtree

• Group unsafe cells, until safe
• Pro: protection garanteed, relative low adaption.
• Con: blocky result

pop_quad <- protect_quadtree(population, sensitive = FALSE)
plot(pop_quad, "count")

Utility measures

Hellinger Distance (HD)

• See Map as a table:
• Difference between unprotected and protected version (rmse diff of normalized data)
• Note: does not take spatial character into account!

Kantorovic-Wasserstein Distance (KWD)

• Spatial distribution distance
• aka: earth movers distance, minimal spatial distribution adjustment
• using R package SpatialKWD
• Note: computational intensive

Areas countries

knitr::include_graphics("fig/areas.png")

Results (NL)

knitr::include_graphics("fig/utility_nl.png")

Discussion

• Using HD sdc-table-like measures, cell removal best, smoothing worse
• Same with spatial KWD!

However...

• current utility measure look at individual differences
• It does not take spatial pattern/shape into account
• Smoothing may create utility, reducing noise.
• So looking into other utility spatial utility measures.

Upcoming changes sdcSpatial 0.7

• Hellinger Distance (HD): distance_hellinger()
• speed improvements:
• raster construction (faster)

• smoothing (much more efficient): from city size to country size.

• tbp in January 2024

Thank you for your attention!

Questions?

Curious?

install.packages("sdcSpatial")

Feedback and suggestions?

https://github.com/edwindj/sdcSpatial/issues

References

edwindj/sdcSpatial documentation built on April 13, 2025, 1:57 a.m.