David wrote a number of papers on Database Forensics in the past that were very interesting and it was clear thart this area has some passion for him. Some of his papers focused on analysing redo and data files for evidence of wrong doing and its clear from David's announcement today that he is developing a product around this space to help people do post breach analysis as he says nothing exists - which to my knowledge, also is true.
Analysing the redo and / or data files - is a good idea if its done out side of the Oracle software as any "use" of the Oracle software to perform breach analysis will also affect the database/data and and in-memory view of the database, in otherwords it ends up like heisenbergs uncertaintly principal. The more you measure the more you will affect the result.
There are some downsides. Reading is not normally recorded in the database other than transiently in memory and also possibly on disk if its captured as part of workload/Statspack type events. The problem for me is that a breach does not necessarily change data or structure. If you want to steal credit cards then read them and write them down. In reality reading credit cards (or indeed any other data) leaves a lot of transient evidence.
As an aside some of what David may be doing can be done with the Oracle software (but certainly not all), for instance LogMiner is a great tool to read archive logs and redo logs as is CDC. The only stipulation would be the need to do the reading and analysys on another database so the primary is not affected.
David announced the tool DDLDUMP on the Oracle-l list today. The post is titled "ddldump" and the tool is closed source and available from http://www.v3rity.com/ddldump.php - (broken link) v3rity. A simple sample run is here:
|