Hiya Mike,

> -----Original Message-----
> From: [EMAIL PROTECTED] 
[...]
> 
> Rick Brown wrote:
> > 
> > [ extranet - how? ]
> > The web app needs to access to an internal
> > Oracle database.  I'm wondering what's the best way to
> > set this up?  My first thought was to replicate the
> > database to the DMZ.
> 
> If this is doable, it is indeed a very good design choice. 
> Replicating the bare minimum to the separate zone, and 
> replicating as little as possible back to the inside 
> (preferably nothing, if possible?) is just about as good as 
> it gets. If oracle can be set up so that the internal DB 
> initiates all of the replicating sessions (sorry, me no 
> oracle guru), it would be _much_ preferable to allowing the 
> extranet DB server initiating sessions to the inside.

This doesn't seem right to me.

IMO, the biggest risk with database-backed web thingies is a compromise
of the database server, resulting in the entire database becoming
available to an attacker. I'm assuming that _any_ database would need to
hold sensitive data, like credit card numbers, to be useable, so even
the bare-minimum replicated database would still be sensitive. Given
that assumption, a compromise of the database server on the DMZ could
easily lead to loss of all the marbles - How well can oracle cope with
root compromise of the underlying OS (I'm guessing not well)?

In terms of design choices, I'd prefer to allow queries through, but to
have strong database security. In other words, the external server can
ask the internal server for prices or stock, but not for credit card
numbers. I do appreciate your concern, which is obviously that live
queries against the core production database is a Bad Thing, so how
about an _internally_ replicated system with a well secured database
containing the bare minimums (including credit card numbers) which can
be queried from the DMZ? That obviates the risk of someone destroying
the core DB, and the only price you pay is the possible use of the
internal server as a second stage jump-off point for attacks. One could
reduce _that_ risk by placing the replicated server in a separate
security zone altogether. (But I'd say it's splitting hairs. Someone
would need to r00t the webserver, use an Oracle query to r00t the
internal database and then (using only oracle traffic) use the internal
server to attack the inside zone - it's getting far fetched).

> 
> > Another thought was reverse proxy but I've never done that
> > and I'm wondering how secure that is.
> 
> You'd have to have a very well-written proxy [1] with very 
> fine-grained access control in order for it to improve 
> security even measurably. I have no idea if such a beast 
> exists (oracle not being my strong side and all). 

I'm with you - the only fine-grained Oracle proxy is a well configured
Oracle database.

> /Mikael

Cheers,

--
Ben Nagy
Network Security Specialist
Mb: TBA  PGP Key ID: 0x1A86E304 
(Looking for work in Geneva)

_______________________________________________
Firewalls mailing list
[EMAIL PROTECTED]
For Account Management (unsubscribe, get/change password, etc) Please go to:
http://lists.gnac.net/mailman/listinfo/firewalls

Reply via email to