Anonymizer Buddy

Detect and redact supported sensitive content in text and structured records. The published alpha provides a deterministic Rust/WebAssembly engine for Node.js and browsers, with explicit initialization and disposal.

Profiles and field policies

The deterministic profiles detect validated email addresses, IPv4/IPv6 addresses, and 13–19-digit Luhn-valid card candidates. They do not automatically detect phone numbers, contextual names, addresses, or credentials. Selected-field JSON policies can target known sensitive fields without relying entirely on free-text detection.

An opt-in OCR text profile handles some noisy extracted text. Contextual person-name detection and multilingual model evaluation remain experimental. No production name-detection or language guarantee is implied.

Redaction and review

The pipeline detects spans, reconciles overlaps, rewrites content, and checks the result against the configured policy. Optional findings help explain changes. A completed verification step means those checks ran; it does not mean arbitrary text is anonymous.

Boundaries

• Detection depends on the selected profile and its documented coverage. Misses and false positives are possible.
• Contextual person-name models are experimental; no production language or accuracy guarantee is available.
• Redacted text does not sanitize source images or PDFs, and does not establish universal anonymity.
• Detects validated email, IPv4/IPv6, and Luhn-valid card candidates. Phone numbers, contextual names, and credentials are not automatic detectors.

Review representative data for missed content and unnecessary removals before choosing a profile. Keep sensitive text and findings out of logs and analytics.

Integration

Initialize with createAnonymizer, call detect, redactText, or redactRecordJson, then dispose the engine. Operations are synchronous and bounded; disposal does not promise memory zeroization. No semantic model is included. See the preview status.