DataSunrise Dynamic Data Masking for Greenplum

Greenplum Dynamic Data Masking by DataSunrise.Imagine that contractors you hired need access to your Greenplum database. The problem is that your database stores client addresses, credit card numbers and that sort of stuff. Static data masking is not an option because making a database “dummy” is pricy and time-consuming process. The solution is dynamic data masking. When masking is employed, DataSunrise obfuscates the database output making it meaningless or unreadable.

Tech Info

Greenplum data masking by DataSunrise is an efficient method of preventing sensitive information exposure when giving database access to third-party. There are a lot of situations when organizations need to share data from their production databases with third party. Here are some examples:

  • A certain organization hires outsourced IT specialists to customize its business system or perform database testing, upgrade etc.
  • Healthcare company provides medical researchers with information on clinical trials.
  • Retail company shares sales information with market researchers.

The point is that in most cases third-party specialists don’t need real data the database contains. An environment mimicking a real production database is enough. The best way to protect sensitive data while transferring it to third-party is to replace it with some neutral data. And the most efficient way to do this is masking.

The important point is that masked data should be consistent enough to support proper functioning of the third party’s application. In simple words, the main purpose of masking is to make sensitive data useless for bad guys while keeping it useful for the ones that should receive it.

Sometimes in such situations static masking is used. It is about providing the third party with a stand-alone copy of real database containing some neutral data instead of sensitive data. While this method is reliable, it could be pricey and time-consuming.

That’s why in most cases dynamic data masking is preferable. Unlike static data masking, dynamic data masking is about obfuscating sensitive data on-the-fly, while transferring it to third party. In this case actual database contents remains intact and only database output is obfuscated.

This is how it works

Greenplum data masking tool by DataSunrise is deployed as a proxy between clients (those third-party specialists) and the Greenplum database.

Clients contact production database through DataSunrise proxy only (any direct access to database is disabled).

DataSunrise intercepts client query, changes it according to existing security policies and redirects modified (“masked”) query to the Greenplum database. Having received “masked” query, the database outputs fake (obfuscated) data instead of real values originally requested by the client. Since no real data leaves the database, this method of data obfuscation is very reliable.

And this is how it looks like

Actual list of customers

For example, here we have a Greenplum table which resembles a list of customers including addresses, emails and credit card numbers. Before exporting this list to a third-party system, we need to obfuscate customers’ personal data.

To do this, some masking rules were created. Along with general-purpose masking methods DataSunrise provides obfuscation algorithms for emails and credit card numbers. So, we’ve created some masking rules to obfuscate data in columns that contain customers' addresses, ZIP codes and credit card numbers. Let’s assume that leaving their names and order numbers unmasked is acceptable.

table masked

And here how the table looks like after the masking is applied. As you can see, data in specified columns were obfuscated and thus, it became useless for potential wrong-doer.

Masking internals

Now let's see how masking looks like in SQL.

Here's the original client query:

SELECT oid,
       *
FROM public.customers
ORDER BY oid ASC
And SQL query the database gets after masking is applied:
SELECT oid,
       public.sales_db."Order",
       public.sales_db."First Name",
       public.sales_db."Last Name",
       CAST('MASKED'::text AS text) AS "Address",
       CAST('MASKED'::text AS text) AS "ZIP",
       public.customers."Email",
       CAST(regexp_replace("Card", '([0-9])(?![0-9]{0,3}($| ))', 'X', 'g') AS character (20)) AS "Card"
FROM public.customers
ORDER BY oid ASC

Related Articles