Supported Entities

Pseudonymize.io uses a combination of ML models, pattern matching, and exact validators to detect sensitive data. Not all entity detection is equivalent; understanding the underlying detector helps you anticipate reliability.

Capability Matrix

Entity	Example	Detector	Deterministic	Reversible
Email	[email protected]	pattern	yes	yes/no
Phone	+49 151 2345678	pattern/library	yes	yes/no
Person	Alice Rossi	ML (NER)	yes	yes/no
Location	Munich	ML (NER)	yes	yes/no
Organization	Acme Corp	ML (NER)	yes	yes/no
IP Address	192.168.1.1	pattern	yes	yes/no
IBAN	DE89370400440532013000	validator	yes	yes/no
Payment Card	4532...	validator (Luhn)	yes	yes/no

Detector Types

1. Pattern: Extremely reliable Regex implementations. Expect near 100% precision and recall for properly formatted strings (like IP addresses).
2. Validator: Regex followed by an algorithmic validation check (e.g. Luhn algorithm for credit cards, checksums for IBANs). Prevents false positives on random 16-digit numbers.
3. ML (NER): Natural Language Processing models (Named Entity Recognition). Highly dependent on context. Precision and recall typically hover between 85% - 95%. Expect some false negatives if a name is completely isolated without surrounding grammatical context.

Custom Entities

If your application uses proprietary identifiers (like internal TK-XXXX support ticket IDs), you can pass custom Regex patterns in the API payload under custom_entities (Coming in v1.2).