DocxAnonymizer-spark
Stand-alone Scala & Java tool to anonymize OOXML Documents (docx). This software helps to make docx documents compliant to General Data Protection Regulation 2016/679 (GDPR).
This project is an extension of DocxAnonymizer-core. Both software perform the same tasks, but DocxAnonymizer-core can work only on a single machine, whereas DocxAnonymizer-spark can work either locally or on a cluster with Apache Spark (e.g. on Amazon EMR). The latter implementation is completely scalable and drastically more efficient especially in processing large documents.
This project was developed as part of the Languages and Algorithms for Artificial Intelligence university course (Master in Artificial Intelligence, Alma Mater Studiorum - University of Bologna).
Authors
Workflow
Given a complex and rich-formatted docx file, the program extrapolates all the text, it anonymizes its content and then it saves the new docx without altering its structure.
Nominatives (i.e. sequences of names and surnames) are replaced with anonymous IDs and multiple occurrences of the same nominatives are replaced with the same ID.
The detection of the nominatives can be either on demand or automatic. In fact, the user can express as input the sequences of names-surname to anonymize; in case these sequences are not given, the program automatically starts searching for nominatives in the document using dictionaries of Italian and English names. A pattern-based approach is adopted to detect nominatives. Further details here.
In brief, the program accepts the following options:
-i <input-file> [the docx input file to anonymize]
-o <output-file> [the docx output file generated, if not expressed given by default]
-s3 <s3bucket> [the S3 bucket in which files are stored, all file paths are relative to the bucket. If s3bucket is not provided, files are supposed to be stored locally]
-m <minimize> [the file with names and surnames to minimize. It must contain one expression per line of the form: "<name1>:<name2>:[...]:<nameN>;<surname>", if not expressed the program will perform automatic detection of nominatives]
-kn <keep-names> [the file with names and surnames to keep unchanged (no minimization). It must contain one expression per line of the form: "<name1>:<name2>:[...]:<nameN>;<surname>"]
-ke <keep-expressions> [the file with those expressions to be kept unchanged (not nominatives)]
-p <parallel> [enable distributed execution]
-d <debug> [increase verbosity]Notes
DocxAnonymizer leverage the docx4j library.