Skip to contents

Anonymize sensitive information in R objects

Source: R/anon.R
anon.Rd

This function anonymizes or redacts sensitive information from various R objects including character vectors, factors, data frames, and lists. It uses pattern matching to find and replace sensitive content, with options for targeted anonymization based on variable names or classes and warnings about approximate matches.

Usage

anon(
  x,
  pattern_list = list(),
  default_replacement = getOption("anon.default_replacement", default = "[REDACTED]"),
  check_approximate = getOption("anon.check_approximate", default = FALSE),
  max_distance = 2,
  df_variable_names = NULL,
  df_classes = NULL,
  check_names = TRUE,
  check_labels = TRUE,
  nlp_auto = getOption("anon.nlp_auto"),
  .self = FALSE,
  .pattern_replacements = NULL,
  .compiled = NULL
)

Arguments

x

The object to anonymize. Can be a character vector, factor, data frame, or list.

pattern_list

A list of patterns to search for and replace. Can include:

  • Named elements where names are replacement values and values are one or more patterns to match

  • Unnamed elements where one or more patterns are replaced with default_replacement This parameter is combined with the global option getOption("anon.pattern_list").

default_replacement

Value to use as the default replacement value when no specific replacement is provided. Default is getOption("anon.default_replacement", default = "\[REDACTED\]").

check_approximate

Logical indicating whether to check for approximate matches using string distance. Default is getOption("anon.check_approximate", default = FALSE).

max_distance

Maximum string distance for approximate matching when check_approximate is TRUE. Default is 2.

df_variable_names

For data frames, a character vector or named list specifying which variable names should be anonymized:

  • Unnamed elements: variables are replaced with default_replacement

  • Named elements: variable names are keys, value can be either a replacement value or a function This parameter is combined with the global option getOption("anon.df_variable_names").

df_classes

For data frames, a character vector or named list specifying which variable classes should be anonymized:

  • Unnamed elements: variables with matching classes are replaced with default_replacement

  • Named elements: class names are keys, value can be either a replacement value or a function This parameter is combined with the global option getOption("anon.df_classes").

check_names

Logical indicating whether to anonymize object names (column names, row names, list names). Default is TRUE.

check_labels

Logical indicating whether to anonymize labels (attributes). Default is TRUE.

nlp_auto

List of logical values with names corresponding to entity names. Can be generated with nlp_auto() and can be set as the anon.nlp_auto global option. This argument overrides the global option.

.self

Logical for internal use only. Used in recursive calls. Default is FALSE. When TRUE, warnings are collected as attributes instead of being issued immediately and global options are ignored and only explicitly provided parameters are used.

.pattern_replacements

List for internal use only. Pre-computed pattern replacement pairs passed down during recursive calls to avoid recomputing them for each column or list element. Default is NULL, which triggers normal computation.

.compiled

List for internal use only. Pre-compiled pattern groups containing grouped regex objects, digest tokens, and fixed token matchers. Passed down during recursive calls to avoid recompiling patterns. Default is NULL.

Value

An object of class anon_context with the same structure as x but with sensitive information replaced. If approximate matches are found and .self is FALSE, warnings are issued. If .self is TRUE, warnings are attached as an attribute.

Details

anon() operates recursively on nested structures. For data frames:

  • Individual columns are processed based on their content

  • Entire columns can be replaced if specified in df_variable_names or df_classes

  • Column names, row names, and labels are anonymized when enabled

Pattern matching is case-insensitive. When check_approximate is enabled, anon() will warn about remaining potential matches that are similar but not exact.

Replacement functions can be provided in df_variable_names and df_classes as:

  • R functions that take the column/variable as input

  • Formula notation (e.g., ~ .x + rnorm(length(.x), mean = 1))

The returned object has class anon_context which allows it to be combined with other anonymized objects using c().

Global Options

The following global options affect function behavior:

anon.default_replacement

Default replacement text (default: "[REDACTED]").

anon.pattern_list

Global patterns to combine with (after) pattern_list parameter.

anon.df_variable_names

Global variable name specifications to combine with (after) df_variable_names parameter.

anon.df_classes

Global class specifications to combine with (after) df_classes parameter.

anon.nlp_auto

List of logical values indicating which NLP entity types should be automatically anonymized. Use nlp_auto() to generate this list. Override the option by setting the nlp_auto argument.

anon.nlp_default_replacements

Default NLP replacement labels. Use nlp_default_replacements() to generate this list.

anon.example_values_n

Default example_values_n used by anon_data_summary() and anon_report().

anon.example_rows

Default example_rows specification used by anon_data_summary() and anon_report(). Use anon_example_rows() to generate this value.

See anon_options() for a central helper that lists and sets all supported anon.* options.

To set global options:

options(anon.pattern_list = list("EMAIL" = "@\\S+"))
options(anon.df_variable_names = c("name", "email"))
options(anon.default_replacement = "[HIDDEN]")
options(anon.nlp_auto = nlp_auto(person = TRUE))

Examples

# Basic string anonymization
text <- c("John Smith", "jane.doe@email.com", "Call 555-1234")
anon(text, pattern_list = c("John Smith", "@\\S+", "\\d{3}-\\d{4}"))
#> [1] "[REDACTED]"         "jane.doe[REDACTED]" "Call [REDACTED]"   

# Using named patterns for specific replacements
anon(text, pattern_list = list("PERSON" = "John Smith",
                               "EMAIL" = "@\\S+",
                               "PHONE" = "\\d{3}-\\d{4}"))
#> [1] "PERSON"        "jane.doeEMAIL" "Call PHONE"   

# Data frame anonymization
df <- data.frame(
  name = c("Alice", "Bob"),
  email = c("alice@test.com", "bob@test.com"),
  score = c(95, 87)
)

# Anonymize specific columns by name
anon(df, df_variable_names = c("name", "email"))
#>         name      email score
#> 1 [REDACTED] [REDACTED]    95
#> 2 [REDACTED] [REDACTED]    87

# Anonymize columns by class with custom replacements
anon(df, df_classes = list("character" = "HIDDEN"))
#>     name  email score
#> 1 HIDDEN HIDDEN    95
#> 2 HIDDEN HIDDEN    87

# Using functions for dynamic replacement
anon(df, df_variable_names = list("name" = ~ paste("Person", seq_along(.x))))
#>       name          email score
#> 1 Person 1 alice@test.com    95
#> 2 Person 2   bob@test.com    87

anon_df <- df |>
  anon(
    df_variable_names = list(
      "name" = ~ paste("Person", seq_along(.x)),
      "email"
    )
  )

# Using global options
options(anon.pattern_list = list("EMAIL" = "@\\S+"))
options(anon.df_variable_names = "name")
anon(df)  # Will anonymize emails and names using global settings
#>         name      email score
#> 1 [REDACTED] aliceEMAIL    95
#> 2 [REDACTED]   bobEMAIL    87

# Combine anonymized objects
anon_summary <- anon_data_summary(list(df = df))
combined <- c(anon_df, anon_summary)
combined
#> === ANONYMIZED DATA CONTEXT ===
#> 
#>       name      email score
#> 1 Person 1 [REDACTED]    95
#> 2 Person 2 [REDACTED]    87
#> 
#> Environment Data Summary
#> ========================
#> 
#>   total_objects data_frames other_objects total_memory
#> 1             1           1             0         1272
#> 
#> Data Frames:
#> ------------
#>   name       type n_rows n_cols memory_size
#> 1   df data.frame      2      3      1.2 Kb
#> 
#> 
#> Variable Details (df):
#> 
#> -------------------- 
#>   variable data_type n_distinct n_missing n_total pct_missing label
#> 1     name character          2         0       2           0  <NA>
#> 2    email character          2         0       2           0  <NA>
#> 3    score   numeric          2         0       2           0  <NA>
#> 
#> 
#>