Pete Finnigan's Oracle Security Forum


    
      
        Pete Finnigan's Oracle Security Forum
        (http://www.petefinnigan.com/forum/yabb/YaBB.cgi)
      

        Oracle Security >> Oracle Security >> The owner of the listener process
        
(Message started by: Pete Finnigan on Jan 29^th, 2008, 3:39pm)

Title: The owner of the listener process
Post by Pete Finnigan on Jan 29^th, 2008, 3:39pm

A current security requirement in my organization is to have the listener owned by a different account than the instance. Can someone support the reasoning of this requirement?

1. Although giving it a different owner is easy enough, the listener owner account has to be in the dba group to access the data files correct?

2. Split out or not, hacking the owner of the listener process still leads to a denial of service. Once hacked, the group privilege could be used to cause other damage. What have we gained by splitting this out?

3. Is having the listener and instance owned by the same account but keeping the account locked not a secure approach? We currently have to sudo to root then su to the owner account. Is that not a safe approach?

Any feedback on this would be appreciated.

Title: Re: The owner of the listener process
Post by Pete Finnigan on Jan 30^th, 2008, 4:03pm

Sorry Pete, I should have posted this to the Oracle Security section.

Title: Re: The owner of the listener process
Post by Pete Finnigan on Jan 31^st, 2008, 11:27am

Hi Cody,

dont't worry its fine, i moved it to here. The reasoning for owning the listener by another user such as "nobody" is that if its hacked then the hacker is nobody not Oracle.

I am unsure why you need to thave the listener process access datafiles? - you mean because the listener daemon spawns the shadow process? - the shadow process is the clients communication channel, it should not need access to datafiles?

Normally this sugestion is made for extproc and Oracle have documented how to do it on a number of occasions on older security alerts. I know of people who have moved apache to apache:apache - there is a paper on this out there but I cannot locate it quickly now. Also some people have the listener running as nobody - that is the daemon is still owned by oracle:dba (ok, i know these are bad name choices) and nobody simply starts the listener. I have seen an Oracle document describing this but again cannot find it this morning quickly. This subject was discussed here previously - http://www.petefinnigan.com/forum/yabb/YaBB.cgi?board=ora_sec;action=display;num=1150219028

cheers

Pete

Title: Re: The owner of the listener process
Post by Pete Finnigan on Feb 1^st, 2008, 12:21am

Some Oracle-specific security guidelines say to use different accounts for the database and Listener. I
too have been puzzled by that requirement. In my company, one person
maintains both the database and the Listener. So if the intention is to
protect from a blackhat admin, that person already has full control. There
is the special condition of extprocs: if extprocs are being used, then yes,
absolutely, that extproc Listener should be owned by a non-privileged
account. The next question is the general case, the Listener handling
normal traffic. One attack vector would be Listener administration: if the
Listener administration is not protected adequately then a blackhat could
direct the log files to overwrite the data files and thus destroy
availability until a restore can take place. On the other hand, the 10G
local authentication solves this issue. The remaining vector would be some
kind of spectacular buffer overflow attack and my gut feel is that, by now,
with current software, those holes should have been closed. My conclusion
is that if (a) it is a 10G or higher database (b) administration of the
Listener is only via local authentication (c) the Listener is not used for
extprocs, then it should be safe to use the same account. I don't know that you could convince whomever it is who mandates security controls, though.

By the way, if the Listener is run by another userid, then that userid should not be in the OSDBA group. You're protecting from somebody who can hack the Listener; if they can do that, they can figure out how to muck with the Listener to use SYSDBA privileges to come into the database. (I wouldn't know how to do it but there seem to be people out there who can.)

Title: Re: The owner of the listener process
Post by Pete Finnigan on Feb 6^th, 2008, 10:11pm

Sorry to reply so slowly. Flu took me out for a bit. This is definitely something our group of DBAs did not know. Thanks for straightening us out.