cryptRopen

cryptRopen pseudonymizes variables in one or many datasets, driven by a single Excel mask that lists, per input file, which columns to hash and which to drop. The transformation is a salted hash (MD5 by default; any algorithm supported by digest::digest() works), which means:

• The same input value paired with the same key always yields the same hash. Joins across tables on the pseudonymized columns stay possible.
• The hash is one-way: holding the output does not let anyone recover the input. Only the correspondence table (kept by the party that owns the key) does.

Suitable for GDPR-style pseudonymization. Not suitable as a reversible encryption scheme for confidentiality-bearing data.

Installation

Install the development version from GitHub:

# install.packages("pak")
pak::pak("danielrak/cryptRopen")

In-session use: crypt_data()

Pseudonymize one data frame already in memory. The correspondence table is kept in a package-private environment and retrieved with get_correspondence_tables():

library(cryptRopen)

pseudonymized <- crypt_data(
  loaded_dataset             = head(mtcars, 3),
  vars_to_encrypt            = "mpg",
  vars_to_remove             = "cyl",
  encryption_key             = "demo-key",
  correspondence_table       = TRUE,
  correspondence_table_label = "demo"
)
pseudonymized
#>                                      mpg_crypt disp  hp drat    wt  qsec vs am
#> Mazda RX4     C33FDD39F6EAADF6C5E135AD5083E3D6  160 110 3.90 2.620 16.46  0  1
#> Mazda RX4 Wag C33FDD39F6EAADF6C5E135AD5083E3D6  160 110 3.90 2.875 17.02  0  1
#> Datsun 710    7E7E52C52B1C3686F553D1E4AC6A8207  108  93 3.85 2.320 18.61  1  1
#>               gear carb
#> Mazda RX4        4    4
#> Mazda RX4 Wag    4    4
#> Datsun 710       4    1

get_correspondence_tables()$tc_crypt_demo
#>                mpg                        mpg_crypt
#> Mazda RX4     21.0 C33FDD39F6EAADF6C5E135AD5083E3D6
#> Mazda RX4 Wag 21.0 C33FDD39F6EAADF6C5E135AD5083E3D6
#> Datsun 710    22.8 7E7E52C52B1C3686F553D1E4AC6A8207

Batch / file-driven use: crypt_r()

crypt_r() reads an Excel mask describing N input files, dispatches one mirai::mirai() task per row (parallel, non-blocking), writes the pseudonymized files and a recap xlsx log to disk, and returns a cryptR_job handle. Inspect it with cryptR_status() / summary() / cryptR_results() and finalize with cryptR_collect().

job <- crypt_r(
  mask_folder_path  = "path/to/mask",
  mask_file         = "mask.xlsx",
  output_path       = "path/to/output",
  intermediate_path = "path/to/intermediate",
  encryption_key    = "your-salt",
  algorithm         = "md5"
)

cryptR_status(job)
job <- cryptR_collect(job)

Each row of the mask declares its own vars_to_encrypt, so different input files can have entirely different column names — the mask is the contract, not a single schema. Large parquet / CSV inputs are automatically routed to a streaming engine (arrow::open_dataset() + chunked write) so memory usage stays bounded regardless of input size.

Learn more

• vignette("cryptRopen") — getting started: the in-session workflow with crypt_vector() / crypt_data(), the difference between pseudonymization and encryption, and how to pick an algorithm.
• vignette("crypt_r-workflow") — full walkthrough of the mask-driven crypt_r() workflow: Excel mask layout, async inspection, streaming, daemons ownership, recap log.
• NEWS.md — change history.

License

MIT © Daniel Rakotomalala.