Skip to content

CloakPII

Moving customer data across borders without leaking PII — or failing a PIPL / PDPA audit — is hard. CloakPII does it in one command.

If your data leaves China or Singapore, you're on the hook for PIPL and PDPA: personal data has to be desensitized, the transfer encrypted, and the paperwork filed. Rolling that yourself means stitching together PII detection, a masking scheme, AES encryption, and a compliance checklist — and getting every one of them right.

CloakPII is a single command-line tool that takes a folder of files, masks (or reversibly tokenizes) the PII inside, encrypts every file with AES-256-GCM, and generates the PIPL / PDPA compliance report — so cross-border data is safe to move and you have the audit trail to prove it.

pip install cloakpii

export CLOAKPII_PASSWORD=...      # keep secrets out of ps / shell history
cloakpii migrate --source ./data --output ./safe \
  --compliance-profile pipl --compliance-report

# before                              # after (masked + encrypted)
name,email,phone                      name,email,phone
Wei,wei@corp.cn,138-1234-5678         W***,w***@c******.cn,138-****-**78

Masked plaintext lands in ./safe/desensitized/, AES-256-GCM copies in ./safe/encrypted/, and a PIPL report (JSON + Markdown) next to them. Restore the whole tree any time with cloakpii decrypt-all.

Why CloakPII

CloakPII Microsoft Presidio Roll-your-own gpg / openssl
PII detection (regex + field names) ✅ ✅ ⚠️ DIY ❌
Chinese IDs and Chinese column names ✅ ⚠️ partial ⚠️ DIY ❌
Irreversible masking ✅ ✅ ⚠️ DIY ❌
Reversible, join-preserving tokenization ✅ ❌ ⚠️ DIY ❌
Built-in AES-256-GCM encryption ✅ ❌ ⚠️ DIY ✅
PIPL / PDPA / GDPR compliance reports ✅ ❌ ❌ ❌
Multi-format (CSV/JSON/Excel/Parquet/XML/TSV/SQLite/text) ✅ ⚠️ text-first ⚠️ DIY ⚠️ opaque blobs
One command, folder in → safe folder out ✅ ❌ ❌ ❌

Presidio is a great detection engine, but it stops at detection — you still bolt on encryption and compliance yourself. gpg encrypts but never sees the PII. CloakPII is the opinionated, end-to-end path for the cross-border-compliance use case.

What this is — and what it isn't

Read before you rely on it

  • Masking (the default) is irreversible — masked values can't be recovered even after decrypt. Use --mode tokenize if you need recoverable, join-preserving data.
  • Compliance output is documentation, not legal sign-off. Have counsel review actual cross-border filings.
  • Detection is not exhaustive — it's regex + column-name keywords. It does not catch bare-digit phone numbers (unless the column name signals PII), 15-digit Chinese IDs, IPv6, free-text names, or PII glued to surrounding letters. Enable the ML backend for names, and spot-check a sample of any new dataset.

Next steps