Anonymise Columns¶
Data > Anonymise columns... (Ctrl+Shift+Y) prepares a file for sharing by masking or scrambling sensitive columns. Add one or more rules, pick a strategy for each, choose where the result goes, and press Apply. An Apply is a single undo step (Ctrl+Z reverts the whole operation at once).
Rules¶
Each rule targets one or more columns and applies a strategy:
| Strategy | What it does | Example |
|---|---|---|
| Hash | A stable code derived from the value. The same value always gives the same code, so the data stays join-able. | a@x.com -> 9f86d081a1b2... |
| Partial mask | Keep the first or last N characters, replace the rest. | 5551234 -> ***1234 |
| Redact | Replace the whole value with a fixed token or an empty (null) cell. | secret -> [REDACTED] |
| Fake | A realistic synthetic value (name, email, city, company, phone, UUID). Deterministic, so duplicates stay consistent. | Alice -> Jordan Lee |
Picking several columns in one mask / redact / fake rule applies the same strategy to each of them.
Hashing: SHA-256 vs BLAKE3¶
Both produce a 256-bit digest written as 64 hexadecimal characters.
- SHA-256 is the widely known industry-standard hash.
- BLAKE3 is a modern hash that is much faster on large files.
For anonymisation either is fine and the result is equally join-able. Pick SHA-256 for familiarity, BLAKE3 for speed on big tables. The choice does not change how the output behaves.
Full vs shortened hash¶
By default Octa writes the full 64-character hash. Turn off "Output full hash" to keep only the first N characters as a shorter, tidier ID. The shorter you make it, the higher the (still small) chance that two different values end up with the same code, so keep more characters when the column has many distinct values.
Salt¶
The optional salt is mixed into every value before hashing. The same value plus the same salt always gives the same result, so:
- duplicate values stay linked (the data is still join-able), and
- a later re-run with the same salt re-joins to the earlier export.
A non-empty salt makes the output non-guessable (it defeats lookup tables on low-cardinality columns like emails). Null and empty cells always pass through unchanged.
Combining columns into one ID¶
Select several columns in a single Hash rule to hash them together into
one new column (a pseudonymous key). For example, hashing first + last into
person_id gives every person a stable code without exposing their name. A
multi-column hash always creates a new column (it never overwrites the sources);
set its name in the rule.
Output¶
Choose where the result goes:
- Replace the columns in place - overwrite the chosen columns.
- Add the result as new columns (keep originals) - append the anonymised
values as new columns (e.g.
emailstays,email_anonis added). - Put a sanitised copy in a new tab - leave the original untouched and open a clean copy.
Command line and assistant¶
The same engine is available as octa --anonymize spec.json data.csv (see the
--anonymize reference) and as the
anonymize MCP / assistant tool.