PII De-identifier

:x Supported Files

Use CSV, Excel (XLSX or XLS), Word (DOCX), PDF, text, Markdown, or log files. Format limits affect the downloaded copy.

For CSV files, a first row that looks like values is treated as data.

• :Excel sheet limits
• :Word content and cleaning
• :PDF output and scanned images

:x Browser Processing

Your file is read and treated in this browser. The tool does not upload it to a server. Your original stays unchanged; you choose when to download the treated copy.

Files and results stay in memory during this visit. Reloading the page clears the current file and result.

:What happens to saved custom rules

:x Excel Files

For Excel files, this tool processes only the first visible sheet. Other sheets are left out of the download.

Check the sheet names shown after you add a file.

:x Word Files

DOCX text is treated while supported formatting and surrounding words are retained.

Text checks include headers, footers, footnotes, endnotes, image descriptions, hyperlinks, and title. Comments, tracked changes accepted as final, author/company fields, content-control data, and document variables are cleaned.

Pictures, embedded files, and charts remain unread in the copy. Inspect those separately.

:x Pdf Files

The download is a new text-only PDF with the original page breaks.

The words “Email” and “or” stay. Original layout, images, attachments, and metadata are omitted; no black rectangles are painted over the original page.

Hidden original text can become visible. Read the complete output.

:Why scanned images are not read

:x Scanned Pdfs

The reader needs a text layer; it does not perform optical character recognition (OCR).

A scanned image alone has no readable text. A scan with a hidden text layer can expose that text in the output; review it before sharing.

:x Treatment Choices

:Redact

:Hashing

:Mask data

:Swap for fake data

:x Redact Values

A whole-column finding treats each populated value in that column. A finding inside text changes only the matched characters and keeps the surrounding sentence.

:x Coded Values

Hashing replaces values with keyed hashes. Matching values receive matching hashes within a run or batch. Hashed information can still be linked to people; this is not a guarantee of anonymity.

:Why hashes change between runs

:x Matching Codes

Each run creates a new random key and does not store it. Matching values receive matching hashes within that run or batch. A later run uses a different key, so its hashes differ. The JSON record does not contain the key.

:x Less Exact Values

This keeps a broader value, such as a phone area code, birth year, part of a ZIP code, or network range. The remaining information can still be identifying in context.

:x Fake Values

Swap for fake data replaces supported values with generated data. Matching inputs receive the same fake data within a file or batch.

Fake data is not available for custom rules. Use Keep, Redact, Hashing, or Mask data for those matches.

:x Detection Rules

For a text finding, Redact changes the matched value and keeps the surrounding words.

Rules look for supported patterns in columns and text. A finding's confidence score describes that rule match, rather than the probability that the complete file is safe to share.

:What detection can miss

:x Detection Limits

An unmatched detail can still identify someone.

The name and organization in this sentence remain. Name/place coverage is limited; prose, descriptions, and identifiers unique to your organization may not match. Custom rules can add patterns. Review the retained context and complete output before sharing.

:x Run Record

The JSON record lists the detectors used, treatments, change counts, input and output file hashes, and processing times. Word-file cleaning is listed when applicable. It does not include matched values or the hashing key.

The record documents a run. It is not signed and is not a certificate that the result is anonymous or safe to share.

:What file hashes establish

:Where file processing happens

:x File Hashes

A SHA-256 hash identifies particular file bytes. A matching hash helps check that the file has not changed. It does not establish that the treatments found every identifying detail.

:x Saved Rules

Saved custom rules remain in this browser's local storage and run on new files until you delete them. That does not save your uploaded file, generated hashes, or current result. The current file and result are held in memory during this visit.

:x Tool Licenses

Nutshell

Expandable explanations

Rough Notation

Visual emphasis

Nutshell license · Rough Notation license