Selected Press Releases


Next Steps
Press Releases

IRI DarkShield Upgrade Uses Machine Learning to Find & Mask PII

New Data Searching, Classification, and Masking Functions Are Added

Melbourne, FL - December 19, 2018 - Innovative Routines International (IRI, The CoSort Company), has released the second generation of its DarkShield software for the discovery, delivery, and deletion of sensitive data in unstructured documents. DarkShield Version 1 was introduced earlier this year to simultaneously find and redact personally identifiable information (PII) in multiple “dark data” files across a network at once. DarkShield Version 2 (V2) adds more masking functions, Machine Learning (ML) to train and use a Named Entity Recognition (NER) model, and the ability to schedule the search and remediation jobs separately or in one pass.

NER models use Natural Language Processing (NLP) technology to reveal sensitive information like people’s names and addresses based on their use in many language contexts -- values that need to be delivered and deleted for GDPR compliance. DarkShield V2 adds semi-supervised model training wizards that use ML to speed the process of building and improving the models. According to DarkShield lead engineer Dmitry Kulakov, “this feature gives users an easy way to train their models so names can be found more reliably in similar kinds of documents.”

NER searches join prior DarkShield V1 search methods -- pattern matching and value lookups. Any combination of search methods can now be used and assigned to data classes defined in DarkShield, which could be credit cards, names, emails, etc. IRI SVP David Friedland explained that data classes also get paired with data masking functions.” This intersection of data search, classification and masking rules in IRI ‘shield’ tools, and the Voracity data management platform that includes them, “supports consistent find/fix operations across all enterprise data sources.”

The list of masking functions are also expanded in DarkShield V2. In addition to redaction, V2 offers encryption, pseudonymization, hashing, encoding, bit scrambling and string manipulation. These functions work interchangeably with the masking functions in IRI FieldShield for databases and flat files, and with IRI CellShield for Excel spreadsheets, so that data masked in one environment can be unmasked in another. Other new features in DarkShield V2 include: the serialization of searching and masking operations so users can modify and schedule their jobs; and, the ability to sense new and changed files to save time in subsequent runs.

DarkShield is a capital expense licensed standalone, or an operational expense inside Voracity. The next release of DarkShield, V3, will support additional unstructured data formats and silos, including various image files.

About IRI, The CoSort Company

IRI is a leading US data management and protection ISV founded in 1978 and represented in 40 cities worldwide. IRI’s uniquely fast and versatile data manipulation engines -- and its free Eclipse job design environment -- provide uniquely price-performant and versatile solution software for big data and BI/DW architects, data security and governance officers, DBAs, and developers. Visit for more information.

Press inquiries to:
Craig Schein (
IRI, The CoSort Company
+1 800-333-SORT, ext. 229