The recent Oracle exploit posted to Bugtraq
(http://www.securityfocus.com/archive/1/431353) is actually an 0day
and has no patch.
The referenced exploit seems to use GET_DOMAIN_INDEX_METADATA with a
TYPE_NAME that references an attacker-defined package with a
(modified?) ODCIIndexGetMeta function.
Your last example uses GET_V2_DOMAIN_INDEX_TABLES, with arguments that
reference an attacker-defined package with a (modified?)
ODCIIndexUtilGetTableNames function.
Is this a surface-level discrepancy, or is your vector substantively
different than the one in the exploit? If these are different, then
is it possible that last week's exploit was actually fixed?
No; the same problem occurs. This is the kind of general problem I'm
speaking about. Most vendors that actually understand security will look for
other bugs in the same functional area if you point out a bug. IMO, my job
as a security vulnerability researcher is to highlight problem areas - i.e.
areas of functionality that are rife with issues. How can Oracle fix one
issue but miss the same flaw two lines later??? In this case though, we're
not just talking about one flaw but several. Really, it is inconceivable,
yet they, somehow, manage to do it.
God forbid that any of our critical national infrastructure runs on this
product.... oops it does :(
And every version from 8 through 9 to 10 release 2 is vulnerable. That's
every supported version of Oracle on every operating system.
Oracle customers: honestly - Oracle are not going to listen to the likes of
me - but they will listen folks like you. If you're not happy with the
response you're getting from Oracle then get on the 'phone - call them up
and tell them that you're not happy. Please, demand improvements.
By the way, this is not an isolated incident. I have many examples to hand
where Oracle have tried to fix problems in the same functional area but only
whitewashed it. They should be proactively looking for similar issues in the
same code just like Microsoft does.
The "champion of quality coding movement"
(http://www.cio.com/archive/031505/security.html) , who "applauds ethical
hacking", asks "Why isn't that standard development process?"
I don't know... but I don't think we'll find out in the two year time frame
posited; we've got less than a year to go.
- Steve
P.S. For those of you who are paying attention at this excruciating
level of detail, it seems that David's original use of
GET_DOMAIN_INDEX_METADATA in 2004 directly included the code in the
NEWBLOCK argument, whereas last week's exploit was performed through
an indirect reference to the code in the TYPE_NAME argument.
p.p.s.
Just to clarify the issues:
GET_DOMAIN_INDEX_TABLES
GET_DOMAIN_INDEX_METADATA
GET_V2_DOMAIN_INDEX_TABLES
are all vulnerable to the exploit.
Cheers,
David Litchfield
NGSSoftware Ltd,
http://www.ngssoftware.com/
+44 (0) 208 401 0070
