Document anonymization

Tabla de contenidos

Athento allows the anonymization of PDF documents to hide sensitive information from users and protect confidential data.

The process of anonymizing PDF documents consists of three steps:

  1. Detection of sensitive entities in the document.
  2. Anonymization of the document based on the detected sensitive entities.
  3. Viewing the anonymized document.

Detection of sensitive entities in the document

To detect sensitive entities in a document, you must run the automation "op_anonymize_pdf_entities", available in the space's automation list. This automation works only with PDF documents.

The automation has two parameters:

  • Black words: entities that must be ignored during detection. In some cases, the operation may identify entities that should not be anonymized. It is recommended to test it with a set of documents and include in this parameter the entities you want to exclude. They must be separated by commas.
  • Extract text: checkbox to extract text from PDF documents that do not have digitized text.

Once the operation has been executed on a PDF document, a feature will be displayed with the sensitive entities detected, indicating on which page and at which position each entity is located.

This value can only be seen by users with administration permissions on the document.

Detection of sensitive entities in the document using presidio

Athento has an advanced entity detection version based on Presidio: the automation "op_anonymize_pdf_entities_v2", available in the space's automation list. Like the basic version, it works only with PDF documents, but it allows selecting the document language, the types of entities to detect, and refining the result using reinforcement and exclusion dictionaries.

This automation has the following parameters:

  • Languages (Languages): document languages. If left empty, a generic multilingual model is used. Selecting the specific language improves detection accuracy.
  • Items to detect (Items to detect): types of sensitive entities you want to detect in the document. If none is selected, by default person names, email addresses, and phone numbers are detected. The available types are: person names, emails, phone numbers, locations, dates, IBAN codes, IP addresses, NIF and NIE.
  • Reinforcement dictionary (Reinforcement dictionary): dictionary whose terms are forced to always be detected, even if Presidio does not recognize them on its own. It is useful for business-specific terms (uncommon names, internal codes, etc.). Terms are grouped by category. Only active terms in the dictionary are taken into account.
  • Exclusion dictionary (Exclusion dictionary): dictionary whose terms are removed from the detection result. Each term can be excluded by its exact value or using a regular expression, which provides more control than a simple word list. It is the equivalent of the "Black words" parameter in the basic version. Only active terms in the dictionary are taken into account.
  • Extract text (Extract text): checkbox to extract text from PDF documents that do not have digitized text, using OCR.
  • Pages (Pages): range of pages of the document that are taken into account for entity detection. You can specify an individual page, a page range using a hyphen as a separator (for example 4-6 for pages 4, 5 and 6), or even a union between multiple ranges separated by a comma (for example "1,4-6" for pages 1, 4, 5 and 6). If left empty, all pages of the document are analyzed.

Once the operation has been executed on a PDF document, a feature will be displayed with the sensitive entities detected, in the same way as in the basic version. This value can only be seen by users with administration permissions on the document.

Anonymization of the document based on the detected sensitive entities

After extracting the sensitive entities from the document, an automation must be executed to hide them within the PDF.

This automation is "op_anonymize_document". When it is executed, the entities classified as sensitive in the previous step will be hidden using black boxes, generating a new document with these redaction marks.

Pseudonymization of the document (reversible alternative)

In addition to anonymization (which hides the data with black boxes irreversibly), Athento allows you to pseudonymize a document: each sensitive datum is replaced by a reversible code, so that whoever has the key can recover the original value, even outside Athento.

Pseudonymization reuses the same entity detection as anonymization. The process is:

  1. Detect the entities with op_anonymize_pdf_entities or op_anonymize_pdf_entities_v2.
  2. Run the op_pseudonymize_document automation. It generates a copy of the PDF (the original binary is not modified) in which each detected entity is replaced by its code, and stores the result in the pseudonymization.pdf feature.

How the pseudonymized document looks:

  • Data that fit in their space are replaced by a code of the same size, in place, inside a box.
  • Very long data are shown with a label [PII-1], [PII-2]…, and the full code appears in the tooltip when hovering the mouse over the box.

Unlike anonymization, the code is not a strikethrough: it is an encrypted value that preserves the format of the original (same length and form) and is reversible with the key.

Note: it is only available for PDF documents, and requires having previously run detection with "Include coordinates".

Revert the data (outside Athento)

Important: anyone who has the key can revert the data, so it must be kept securely. The same value always generates the same code within the document (allows recognizing repetitions). If the PDF is printed or flattened, the long codes (which are in the tooltip) may be lost; the short ones are preserved.

Viewing the anonymized document

Once the anonymized or pseudonymized document has been generated, we can make that version be the one shown by default. To do this, in the form configuration section, the option "Default binary version" allows choosing between three values:

  • None: the original binary is shown.
  • Anonymized: the anonymized version is shown, if it exists.
  • Pseudonymized: the pseudonymized version is shown, if it exists.

Once this configuration is selected, whenever we have an anonymized binary generated for our document, it will be shown to all users. Users who are superusers will have the option to view the original binary by pressing the following button.

Users without superuser permission will always see the anonymized document.

**Note: This is only available for PDF documents