Dataset Pseudonymizer

Replace selected CSV values with random tokens or repeatable keyed HMAC pseudonyms.

On-device No sign-up Free to use

Inspect input controls

Showing an example. Edit to see your own.

?

16–4,096 UTF-8 bytes. Keep it separate from the shared dataset. Same key, exact column name and value produce the same token. No key is downloaded.

?

Up to 5 MiB, 20,000 data rows, 50 columns and 200,000 data cells. Every row must have the same width.

Drop your CSV file here

or choose one from your device

One file · CSV and TSV only

    ?

    One exact header name per line. Case and spaces matter. Every selected column gets its own mapping; equal original values in different columns receive different tokens.

    ?

    Anyone with this file can reverse these substitutions. Keep it apart from the shared dataset. Pseudonyms do not make data anonymous.

    ?

    Prefixes formula-triggering exported cells, including unselected cells, with an apostrophe. This changes downloaded values. Check import behavior in the receiving spreadsheet; the preview retains the original text.

    Processed in your browser. Your inputs stay on this device.

    Result

    Showing an example.

    Replace the two repeated names and emails with per-column random tokens while retaining three rows.

    Replaced 4 distinct column values in 3 rows.

    Loading the tool. Your result will appear here.

    https://usefreetools.net/dataset-pseudonymizer · 20 November 2026

    Was this helpful?

    How to use Dataset Pseudonymizer

    1. Paste a CSV with a distinct header row, or choose Upload UTF-8 CSV and select one file.
    2. Enter the exact header names to replace, one per line. Case and spaces matter; blank or whitespace-only values remain unchanged.
    3. Choose fresh random mapping for this run, or Repeatable HMAC-SHA256 mapping with a private key for consistent tokens across runs. Keep that key separate from shared data.
    4. Review the preview, including unselected columns. Download the transformed CSV and enable the sensitive mapping download only when you have a separate protected place for it. Check prefixed cells in the receiving spreadsheet.

    Example: Dataset Pseudonymizer

    Replace the two repeated names and emails with per-column random tokens while retaining three rows.

    You add
    Pseudonym method: random; CSV input: text; CSV with a header row: name,email,account Ada,ada@example.com,0012 Ben,ben@example.com,0034 Ada,ada@example.com,0012; Column names to pseudonymize: name email; Also download the sensitive original-to-token mapping: false; Guard CSV spreadsheet formulas on download: true
    You get
    Replaced 4 distinct column values in 3 rows.

    Options

    Pseudonym method
    Choose Fresh random mapping per run, Repeatable HMAC-SHA256 mapping.
    Private mapping key (keyed mode)
    16–4,096 UTF-8 bytes. Keep it separate from the shared dataset. Same key, exact column name and value produce the same token. No key is downloaded.
    CSV input
    Choose Paste CSV, Upload UTF-8 CSV.
    CSV with a header row
    Up to 5 MiB, 20,000 data rows, 50 columns and 200,000 data cells. Every row must have the same width.
    CSV file
    Choose the local file for Dataset Pseudonymizer.
    Column names to pseudonymize
    One exact header name per line. Case and spaces matter. Every selected column gets its own mapping; equal original values in different columns receive different tokens.
    Also download the sensitive original-to-token mapping
    Anyone with this file can reverse these substitutions. Keep it apart from the shared dataset. Pseudonyms do not make data anonymous.
    Guard CSV spreadsheet formulas on download
    Adds an apostrophe before formula-triggering exported cells, including unselected cells. This changes the downloaded value while the preview retains its text. Review import behavior in the receiving spreadsheet; the prefix is not a guarantee that arbitrary CSV content is safe.

    Supported inputs and limits

    • Up to 5 MiB of CSV, 20,000 data rows, 50 columns and 200,000 data cells.
    • The preview shows at most 100 rows;
    • CSV output includes all accepted rows.
    • Random tokens are consistent within a selected column and run, but change across runs.
    • Keyed mode uses HMAC-SHA256 with a 16–4,096-byte UTF-8 key over the exact column name and value; consistent tokens allow linkage.
    More details

    The key is not exported.

    Optional mappings reveal originals.

    Prefixing formula-triggering cells can change unselected exported values and requires review in the destination spreadsheet.

    Pseudonymization does not make data anonymous.

    Where your input is processed

    This tool does not upload your inputs. See the limits for anything saved on your device.

    Pseudonyms are not anonymity

    A stable token makes repeated records easier to compare, but unselected dates, account numbers or unusual combinations can still identify a person. Inspect the whole table before sharing. The mapping, if requested, directly links original values to tokens and must not travel with the supposedly protected dataset.

    Choose whether tokens must match across runs

    Fresh random mapping is useful for a one-off transformed copy; a new run produces different tokens. Keyed mode supports consistent substitutions across files using the same private key and exact column names and values. Changing name case or spaces changes that input to the HMAC. Consistent tokens let recipients link records, so review whether that linkage is intended before sharing. Neither mode removes identifying detail in other columns.

    Questions about Dataset Pseudonymizer

    Is the output anonymous?

    No. Other columns, patterns or an exported mapping can still identify people.

    Do repeated identifiers receive the same replacement?

    Within a selected column, equal exact nonblank values share one token. Random mappings change on a new run. Keyed mode produces the same token across runs only when the private key, exact column name and original value are the same.

    Should I share the mapping with the dataset?

    Only when intended recipients need it. The mapping reconnects replacements to original identifiers.

    Page updated · Project managed by Tony Hines · Report a problem with Dataset Pseudonymizer