Re: How to make data unidentifiable?

"Joseph" <j_parrinello_at_msn.com> wrote in message news:dd497a10.0406100503.1291ca15_at_posting.google.com...
> Sybrand Bakker <sybrandb_at_hccnet.nl> wrote in message
news:<2stec0ptn1k69cis0oflalr79osnkcttdj_at_4ax.com>...
> > On 9 Jun 2004 10:00:49 -0700, j_parrinello_at_msn.com (Joseph) wrote:
> >
> > >Hello,
> > >
> > >I am building a database to be used for demonstration purposes. I wish
to
> > >import the data from an existing database into the demo. However I
need to
> > >tranform the data from the existing database so that the data is
> > >unidentifiable but meaningful. For example, I need to change the
First
> > >name, Last name, Address, city, state, zip, Social Security etc...so
that
> > >they are different to the original database but still have meaning in
the
> > >new database. E.g. A record such as:
> > >
> > >Peter Paulas, 123 Main St, New York, New York, 10023, 555-342-3423
> > >becomes something like:
> > >Simon Trickle, 234 East St, Bronx, New York, 10023, 534-234-2333
> > >
> > >I would like the transformed record: Simon Trickle, to still be linked
to
> > >the remaining data that is not tranformed. Has anybody tried this
before
> > >and if so, do you have any easy solution.
> > >
> > >Thanks,
> > >Joseph
> >
> >
> > Oracle supplies a dbms_obfuscation_toolkit for this very purpose.
>
>
> Thanks for the input. Even though I have never used the Obfuscation
> toolkit, it appears that Obfuscation simply encrypts the data.
> Instead, I want to transform the data (i.e. only some fields in all
> rows), so that the transformed data can no longer be indentifiable to
> an actual person, however the data still has meaning. So say for
> example:
>
> sql> desc name
>
> Name
> ----------------------------
> ID CHAR(12)
> FIRST_NAME VARCHAR2(30)
> LAST_NAME VARCHAR2(30)
> SOCIAL_SEC CHAR(9)
>
> After I "transform" the data in this table, for each row in the table,
> the ID stays the same. However the FIRST_NAME, LAST_NAME and
> SOCIAL_SEC would be changed to a different FIRST_NAME, LAST_NAME and
> SOCIAL_SEC. Simply, jumbling the letters or encrypting each field
> won't do because I want the data to remain meaningful (i.e. the
> transformed FIRST_NAME is still a real first name). Does this make
> sense? Maybe what I am trying is impossible??

Doesn't make much sense.. only solution i can think of is to have a translation table that translates a name X to another one and that you automatically update by picking a new_name for an new name from the list of existing names in a random way.

Why would you like to "mix up the users" anyway? :)

• Sten