Nearest-Neighbor Distance Ratio privacy score

For each synthetic row, computes the ratio of its distance to the nearest real row vs. its distance to the second-nearest real row. A high ratio (close to 1) means the synthetic row is not unusually close to any specific real row — low disclosure risk. Score = mean(ratio > 0.5).

Usage

nndr(real, synthetic, normalize = TRUE)

Arguments

real, synthetic

Data frames; only numerical columns are used.

normalize

Logical. When TRUE (default), columns are z-scored using the real-data mean and standard deviation before distance computation. Constant columns in real are dropped to avoid division by zero.

Value

A scalar score in [0, 1]; higher = more private.

Details

By default columns are z-scored using the real-data mean and standard deviation before the Euclidean distance is computed; without this, a single large-scale column (e.g. income in dollars) dominates the distance and the score becomes a function of measurement units rather than of similarity.

Examples

real <- data.frame(x = rnorm(50), y = rnorm(50))
syn  <- data.frame(x = rnorm(50), y = rnorm(50))
nndr(real, syn)
#> [1] 0.84