Pseudonymize Variables in a Data Frame

In-session version of the pseudonymization workflow: takes a data frame, hashes the requested columns with a user-provided salt, and returns the transformed data frame. The hash is deterministic and one-way (see crypt_vector() for the details of the transformation).

Usage

crypt_data(
  loaded_dataset,
  vars_to_encrypt,
  vars_to_remove = NULL,
  encryption_key,
  algorithm = "md5",
  correspondence_table = TRUE,
  correspondence_table_label = NULL
)

Arguments

loaded_dataset

A data frame.

vars_to_encrypt

Character vector of column names to hash. Must resolve to at least one non-empty entry after trimming and dropping NA / blank values; an empty vector raises an error.

vars_to_remove

Character vector of column names to drop from the output. NULL (default) keeps all non-hashed columns.

encryption_key

Character scalar. The salt prepended to each value before hashing. See crypt_vector().

algorithm

Character scalar. Any algorithm accepted by the algo argument of digest::digest(). Defaults to "md5".

correspondence_table

Logical scalar. If TRUE (default), stores the correspondence table in the package-private environment under the name tc_crypt_<correspondence_table_label>. Set to FALSE to skip the side effect entirely.

correspondence_table_label

Character scalar. Required when correspondence_table = TRUE. Used to build the storage name so successive calls in the same session do not overwrite each other.

Value

A data frame with the same row order as loaded_dataset, the vars_to_encrypt columns replaced by their hashed counterparts (suffix _crypt), and the vars_to_remove columns dropped.

Details

When correspondence_table = TRUE, the (original, hashed) pairs are stored in a package-private environment under the name tc_crypt_<correspondence_table_label> and retrieved via get_correspondence_tables(). No file is written; this side effect replaces an older behavior that polluted globalenv().

crypt_data() is fail-fast on an empty vars_to_encrypt: it raises an explicit error rather than silently returning the input unchanged. For a "drop columns / convert format" workflow on a batch of files described by an Excel mask, use crypt_r() with a blank vars_to_encrypt cell — that path is legitimate. On a single in-memory object, use dplyr::select() / rio::export() directly.

See also

crypt_r() for the batch / file-driven counterpart, get_correspondence_tables() to retrieve the stored mapping, crypt_vector() for the underlying vector transformation.

Examples

# Replace `mpg` with `mpg_crypt`, drop `cyl`, no correspondence
# table side effect on this run.
crypt_data(
  loaded_dataset       = mtcars[1:5, ],
  vars_to_encrypt      = "mpg",
  vars_to_remove       = "cyl",
  encryption_key       = "1234567",
  algorithm            = "md5",
  correspondence_table = FALSE
)
#>                                          mpg_crypt disp  hp drat    wt  qsec vs
#> Mazda RX4         CEA9D4F43791ACE22D90EDBF6CE405F5  160 110 3.90 2.620 16.46  0
#> Mazda RX4 Wag     CEA9D4F43791ACE22D90EDBF6CE405F5  160 110 3.90 2.875 17.02  0
#> Datsun 710        65080BE2078D5C5934F018F806E72E81  108  93 3.85 2.320 18.61  1
#> Hornet 4 Drive    18C5941010E3CDCB669F472DF3A1DD5E  258 110 3.08 3.215 19.44  1
#> Hornet Sportabout 1866EA523631A281BD9431915F7C407F  360 175 3.15 3.440 17.02  0
#>                   am gear carb
#> Mazda RX4          1    4    4
#> Mazda RX4 Wag      1    4    4
#> Datsun 710         1    4    1
#> Hornet 4 Drive     0    3    1
#> Hornet Sportabout  0    3    2