Pii Detector

A package to handle the detection of PII in text.

TextAnalyzer(additional_recognizers=None, nlp_configuration=None)

Bases: BaseTextAnalyzer

TextAnalyzer class to analyze the text and extract the PII entities.

Implemented using presidio pii library.

Additional recognizers with async def analyze(...) are wrapped so the public analyze method remains synchronous.

Attributes:

Name	Type	Description
analyzer	AnalyzerEngine	The analyzer engine to analyze the text and extract the PII entities.

Initialize the TextAnalyzer class and add custom recognizer.

The predefined recognizers are BankAccountNumberRecognizer, BPJSNumberRecognizer, IndonesianPhoneNumberRecognizer, GDPLabsEmployeeIdRecognizer, FacebookAccountRecognizer, FamilyCardNumberRecognizer, KTPRecognizer, LinkedinAccountRecognizer, MoneyRecognizer, NPWPRecognizer, OrganizationNameRecognizer, and ProjectNameRecognizer. We also add Presidio EmailRecognizer and PhoneRecognizer for Indonesian language.

Parameters:

Name	Type	Description	Default
additional_recognizers	Sequence[EntityRecognizer] | None	The list of additional recognizers to be added. Default is None. Async recognizers are wrapped internally to support sync analysis flow.	None
nlp_configuration	dict[str, Any] | None	Retained for API compatibility and ignored because the analyzer always uses the model-free NLP engine.	None

analyze(text, language='id', entities=None, score_threshold=0.46, allow_list=None, allow_list_match='exact', regex_flags=re.DOTALL | re.MULTILINE | re.IGNORECASE, **kwargs)

Analyze the text and extract the PII entities.

The method intentionally stays synchronous for backward compatibility. If an additional recognizer returns an awaitable result, it is resolved through an internal background event loop.

Note

This method blocks the caller thread. In async application flows, call this method from a worker thread.

Parameters:

Name	Type	Description	Default
text	str	The text to be analyzed.	required
language	str	The language of the text to be analyzed. Default is "id".	'id'
entities	list[str] | None	The list of entities to be extracted. If you need to extract other entities, add the recognizer in the constructor. Default is None.	None
score_threshold	float | None	The minimum score threshold for the extracted entities. Default is 0.46.	0.46
allow_list	list | None	List of words that the user defines as being allowed to keep in the text. Default is None.	None
allow_list_match	str | None	The matching strategy for the allow list. Default is "exact".	'exact'
regex_flags	int | None	The regex flags for the text analysis.	DOTALL | MULTILINE | IGNORECASE
**kwargs	Any	Additional keyword arguments that may be needed for the text analysis process.	{}

Returns:

Type	Description
list[RecognizerResult]	list[RecognizerResult]: The list of extracted entities

TextAnonymizer(text_analyzer, operators=None, add_default_faker_operators=False, faker_seed=42, anonymizer_engine=None, deanonymizer_mapping=None, skip_format_duplicates=False)

Bases: BaseTextAnonymizer

TextAnonymizer class to anonymize the text based on the analyzer results.

Implemented using presidio pii library [1].

Attributes:

Name	Type	Description
anonymizer	AnonymizerEngine	The anonymizer engine used in the anonymization process.
operators	dict[str, OperatorConfig]	The operators used in the anonymization process.
anonymizer_mapping	MappingDataType	The anonymizer mapping.
deanonymizer_mapping	MappingDataType	The deanonymizer mapping.

Initialize the TextAnonymizer class.

Parameters:

Name	Type	Description	Default
text_analyzer	BaseTextAnalyzer	The analyzer engine used in the anonymization process	required
operators	dic[str, OperatorConfig] | None	The operators used in the anonymization process see https://microsoft.github.io/presidio/anonymizer/#built-in-operators	None
add_default_faker_operators	bool	Whether to add default faker operators. Defaults to False.	False
faker_seed	int | None	The seed used for faker. Defaults to 42.	42
anonymizer_engine	AnonymizerEngine	The anonymizer engine used in the anonymization process. Defaults to None.	None
deanonymizer_mapping	DeanonymizerMapping	The deanonymizer mapping used to record mapping between anonymized and original text. Defaults to None.	None
skip_format_duplicates	bool	Whether to skip formatting duplicated operators. Defaults to False.	False

anonymizer_mapping property

Return the anonymizer mapping.

Returns:

Name	Type	Description
MappingDataType	MappingDataType	The anonymizer mapping.

deanonymizer_mapping property

Return the deanonymizer mapping.

Returns:

Name	Type	Description
MappingDataType	MappingDataType	The deanonymizer mapping.

add_operators(operators)

Add operators to the anonymizer.

Parameters:

Name	Type	Description	Default
operators	dict[str, OperatorConfig]	Operators to add to the anonymizer.	required

anonymize(text, entities=None, language='id', conflict_resolution=None, analyzer_score_threshold=0.46, allow_list=None, allow_list_match='exact', regex_flags=re.DOTALL | re.MULTILINE | re.IGNORECASE, **kwargs)

Anonymize the text based on the analyzer results.

Parameters:

Name	Type	Description	Default
text	str	The text to be anonymized	required
entities	list[str] | None	The list of entities to be anonymized. Defaults to None.	None
language	str	The language of the text to be analyzed, Defaults to "id".	'id'
conflict_resolution	ConflictResolutionStrategy | None	The conflict resolution strategy. Defaults to None.	None
analyzer_score_threshold	float	The score threshold for the analyzer. Defaults to 0.46.	0.46
allow_list	list | None	List of words that the user defines as being allowed to keep in the text. Defaults to None.	None
allow_list_match	str | None	The matching strategy for the allow list. Defaults to "exact".	'exact'
regex_flags	int | None	The regex flags for the text analysis.	DOTALL | MULTILINE | IGNORECASE
**kwargs	Any	Additional keyword arguments that may be needed for the text analysis process.	{}

Returns:

Name	Type	Description
str	str	The anonymized text.

deanonymize(text)

Deanonymize the text based on the anonymizer mapping.

Parameters:

Name	Type	Description	Default
text	str	The text to be deanonymized.	required

Returns:

Name	Type	Description
str	str	The deanonymized text or the original text if no mapping is found or the mapping is empty.

remove_all_operators()

Remove all operators from the anonymizer.

reset_deanonymizer_mapping()

Reset the deanonymizer mapping.

anonymize_with_mapping(text, anonymizer_mapping)

Anonymize text with an existing mapping without running PII detection.

anonymizer_mapping groups replacements by entity-type string, such as "PERSON". See :class:Entities for the supported entity types and their matching rules.

Example

anonymizer_mapping = {
    "PERSON": {
        "John Doe": "<PERSON_1>",
    },
}
result = anonymize_with_mapping("Hello John Doe", anonymizer_mapping=anonymizer_mapping)
assert result == "Hello <PERSON_1>"

Parameters:

Name	Type	Description	Default
text	str	Text whose original values should be replaced.	required
anonymizer_mapping	dict[str, dict[str, str]]	An original-to-anonymized mapping in the form {entity_type: {original_value: anonymized_value}}. Entity-type keys must be Presidio entity strings.	required

Notes

Matching is case-insensitive. Most entities use word boundaries; PHONE_NUMBER uses phone-number boundaries instead. PROJECT and ORGANIZATION matches inside URLs are preserved.

Longer originals take precedence over overlapping shorter originals. Equal-length matches prefer URL-preserved entity types, then lexical order, so mapping insertion order does not affect the result. Every match is found in the original input; anonymized output is never processed by another mapping entry.

Returns:

Name	Type	Description
str	str	Text with matching original values replaced by their anonymized values.