IRI, The CoSort Company

In the process of masking data or producing useful test data, you need output target values that look real, but do not reveal personally identifying information (PII). Encryption, de-identification, and other masking, hashing, and bit-level obfuscation functions protect data at risk, but do not provide the level of realism certain recipients require.

Therefore, you need a tool that can change the individualizing characteristics of data in a field by replacing its value with a pseudonym, or substitute (but realistic looking) output value.

One of the challenges specific to pseudonymization in the context of HIPAA compliance is to make sure that the real name cannot be readily discovered through reversal or guesswork.

If you're working with data in tables or flat files, you can use either IRI's FieldShield product or the SortCL interface in your CoSort package to replace personal data with safe, but credible replacement output.

You can specify either an anonymous pseudonym from a random selection of real or fake names (unrecoverable), or specify the use of a set file where real and fake names are either pre-associated, or randomly associated (recoverable).

This functionality -- analogous to a multi-dimensional database lookup table -- can produce a more realistic substitute view of the data than other protection methods.

With pseudonymous data, you can choose to process, share, and even re-identify the data when there is access to the original data and lookup parameters. In the FieldShield context, pseudonymization is also only one method of de-personalizating the information in a record, and can be combined with other data security functions at the field level.

In addition to masking data, there is also a standalone solution for creating realistic test data. IRI's RowGen product uses the same metadata as the FieldShield and CoSort SortCL programs to produce and format test data using pseudonyms. This product is especially useful if you need to provide anonymous, but real-looking, data in database table, flat-file, or report formats, but cannot use real data for testing.