PDF Redaction Tool with Automatic PII Detection and Audit Trail (Desktop)

Sharing documents is routine — offer letters, onboarding records, medical forms, legal filings — and most of them carry Aadhaar numbers, PAN numbers, phone numbers, email addresses or bank details. Drawing a black rectangle over text in a PDF editor does not remove it: the text layer usually survives underneath, and anyone who copies the page can read what was supposedly hidden. True redaction deletes the text objects themselves and paints the region opaque. Meanwhile, compliance workflows need proof of what was done: who opened the file, what was found, what was removed, and when. The available answers are cloud services that require uploading sensitive documents to someone else's servers, or expensive enterprise suites. There is no simple, offline, student-buildable desktop tool that combines transparent PII detection with genuine redaction and a verifiable audit trail — that is the gap this project fills.

1. The tool opens a PDF locally and extracts the text layer page by page — no upload, no network call.
2. The rule engine runs all enabled regex and dictionary patterns over the text, producing a finding list with category, matched value and page location.
3. The document view renders the page with findings highlighted in amber; the inspector lists each finding with its rule, value and field label.
4. On approval, the redaction pass deletes each finding's underlying text objects and paints opaque black regions in their place.
5. The preview view shows redaction coverage before export; the buyer confirms, then the export step writes the redacted PDF with a hash manifest.
6. The audit module logs open, scan, redaction and export events into the hash-chained table, and the buyer verifies correctness with a copy-paste extraction test on the exported file.

Tech stack:

• Python 3.10+
• PyMuPDF (PDF text extraction and object removal)
• Regular expressions + offline name dictionary (detection engine)
• SHA-256 hash-chained audit log
• Electron (desktop shell) or Tkinter/PyQt (alternative UI)
• HTML5 + CSS3 + JavaScript (reference demo UI)
• Synthetic sample PDFs (SAMPLE PII test data)
• Git

• [Rule-based PII detection] Seven transparent rules find emails, Indian mobile numbers, Aadhaar-style numbers (with Verhoeff checksum validation), PAN-style identifiers, bank account numbers, IFSC codes and person names — every rule is inspectable, no black-box model.
• [True irreversible redaction] The tool deletes the underlying PDF text objects and paints opaque black regions, so copy-pasting from the exported file recovers nothing — verified with a before/after extraction check.
• [Redaction preview with coverage] Before exporting, the preview view shows exactly which regions are blacked out and lists every redacted finding by category, page and field label.
• [Hash-chained audit trail] Every document-open, scan, redaction and export event is appended to a tamper-evident log, each entry SHA-256-hashed and chained to the previous one so silent edits are detectable.
• [Rule management view] All seven detection rules are listed with their patterns, categories and actions, each individually toggleable so the buyer can see and control exactly what gets flagged.
• [Synthetic test documents] Ships with sample PDFs containing clearly-marked SAMPLE PII (16 findings across the 2-page demo document), so testing and the viva demo need no real personal data.
• [Export with hash manifest] The export step writes the redacted PDF together with a hash manifest that ties the file to its audit-log entries.

• Working desktop PDF redaction tool (open, scan, preview, redact, export)
• Detection rule engine: 7 documented regex + dictionary rules with Verhoeff validation for Aadhaar-style numbers
• Hash-chained audit-trail module with export manifest
• Synthetic sample PDFs with clearly-marked SAMPLE PII for testing and demos
• Reference single-file HTML/CSS/JS demo of the full UI
• Project report PDF (background, redaction theory, rule design, methodology, verification procedure)
• PPT presentation for final review
• Viva Q&A preparation document (regex design, PDF internals, true vs fake redaction, audit-log hashing)

Timeline

• Built-to-order delivery in approximately 2–3 weeks from order confirmation (expected), covering the build, rule tuning, testing, documentation and the viva kit.

Support

• Approximately 30 days of post-delivery support over WhatsApp and email for installation help, demo walkthrough and viva preparation (expected scope).