Encrypting data for testing - but keeping it legible and the right length

Question

Hello,
I currently work with a client where one of our final stages of testing is done against test databases (9.2), maintained by the client, but holding real data, taken as a snapshot. This gives us volumes of data similar to what will be encountered in production.
The client now has a (valid) concern that our people testing the app therefore get to see 'real', if slightly out-of-date, data, particularly names and addresses, and have asked us if there's some way of hiding this from the testers, who have access to the test db both through a front-end mainly built on Oracle Forms, but also e.g. querying directly with SQL Plus.
I can't just give all the parties stored in the system the same name or address, so what I think I'd like to do is to replace the sensitive data with encrypted versions, possibly using the DBMS_OBFUSCATION_TOOLKIT. A problem with that, though, is that the encrypted data could contain any characters and may cause problems displaying in the front-end, certainly where testers have to e.g. type in a name to search on.
I could use rawtohex to convert the data into an easily read / typed format, but then the results could be too long for the existing columns and the fields in the front-ends.
Has anyone encountered a similar problem and come up with a solution?
Thanks in advance,
James

Answer 1

There are a variety of tools out there sold by companies to automate this sort of thing.
One relatively common, and relatively simple, approach would be to run a process that mixes and matches data from different records. That is, you take the first name from record 1, the last name from record 2, address from record 3... Of course, you'd have to figure out how sophisticated a jumbling algorithm you need depending on how sensitive the data and how performant an algorithm you need depending on data volumes.
Relatively efficient, but relatively simple to reverse, would be to use something like this to populate the tables
insert into emp( ename, sal, comm )
select ename,
      lag(sal) over( order by empno ),
      lag(comm,2) over (order by empno )
  from emp@prod_db_linkJustin