Dear security community and Oracle users,
Many of my customers run Oracle. Much of the U.K. Critical National
Infrastructure relies on Oracle; indeed this is true for many other
countries as well. I know that there's a lot of private information
about me
stored in Oracle databases out there. I have good reason, like most of
us,
to be concerned about Oracle security; I want Oracle to be secure
because,
in a very real way, it helps maintain my own personal security. As
such, I
am writing this open letter
Extract from interview between Mary Ann Davidson and IDG
http://www.infoworld.com/article/05/05/24/HNoraclesecurityhed_1.html
IDGNS: "What other advice do you have for customers on security?"
Davidson: "Push your vendor to tell you how they build their software
and ask them if they train people on secure coding practices. "
Now some context has been put in place I can continue.
On the 31st of August 2004, Oracle released a security update (Alert
68
[http://www.oracle.com/technology/deploy/security/pdf/2004alert68.pdf])
to address a large number of major security flaws in their database
server product. The patches had been a long time in coming
[http://www.eweek.com/article2/0,1759,1637213,00.asp] and we fully
expected that these patches would actually fix the problems but,
unfortunately this is not the case. To date, these flaws are still not
fixed and are still fully exploitable. I reported this to Oracle a
long time ago.
The real problem with this is not that the flaws Alert 68 supposedly
fixed
are still exploitable, but rather the approach Oracle took in
attempting to
fix these issues. One would expect that, given the length of time they
took
to deliver, these security "fixes" would be well considered and
robust;
fixes that actually resolve the security holes. The truth of the
matter
though is that this is not the case.
Some of Oracle's "fixes" simply attempt to stop the example exploits I
sent
them for reprodcution purposes. In other words the actual flaw was not
addressed and with a slight modification to the exploit it works
again. This
shows a slapdash approach with no real consideration for fixing the
actual
problem itself.
As an example of this, Alert 68 attempts to fix some security holes in
some
triggers; the flaws could allow a low privileged user to gain SYS
privileges
- in other words gain full control of the database server. The example
exploit I sent to Oracle contained a space in it. Oracle's fix was to
ignore
the user's request if the input had a space. What Oracle somehow
failed to
see or grasp was that no space is needed in the exploit. This fix
suggests
no more than a few minutes of thought was given to the matter. Why did
it
take 8 months for this? Further, how on earth did this get through QA?
More,
why are we still waiting for a proper fix for this?
Here is another class of thoughtless "fix" implemented by Oracle in
Alert
68. Some Oracle PL/SQL procedures take an arbitrary SQL statement as a
parameter which is then executed. This can present a security risk.
Rather
than securing these procedures properly Oracle chose a security
through
obscurity mechanism. To be able to send the SQL query and have it
executed
one needs to know a passphrase. This passphrase is hardcoded in the
procedure and can be extracted with ease. So all an attacker needs to
do now
is send the passphrase and their arbitrary SQL will still be executed.
In other cases Oracle have simply dropped the old procedures and added
new
ones - with the same vulnerable code!
I ask again, why does it take two years to write fixes like this?
Perhaps
the fixes take this long because Oracle pore through their code
looking for
similar flaws? Does the evidence bear this out. No - it doesn't. In
those
cases where a flaw was fixed properly, we find the same flaw a few
lines
further down in the code. The DRILOAD package "fixed" in Alert 68 is
an
example of this; and this is not an isolated case. This is systemic.
Code
for objects in the SYS, MDSYS, CTXSYS and WKSYS schemas all have flaws
within close range of "fixed" problems. These should have been spotted
and
fixed at the time.
I reported these broken fixes to Oracle in February 2005. It is now
October
2005 and there is still no word of when the "real" fixes are going to
be
delivered. In all of this time Oracle database servers have been easy
to
crack - a fact Oracle are surely aware of.
What about the patches since Alert 68 - the quarterly Critical Patch
Updates? Unfortunately it is the same story. Bugs that should have
been
spotted left in the code, brand new bugs being introduced and old ones
reappearing.
This is simply NOT GOOD ENOUGH. As I stated at the beginning of this
letter,
I'm concerned about Oracle security because it impinges upon me and my
own
personal security.
What is apparent is that Oracle has no decent bug
discovery/fix/response
process; no QA, no understanding of the threats; no proactive program
of
finding and fixing flaws. Is anyone in control over at Oracle HQ?
A good CSO needs to more than just a mouthpiece. They need to be able
to
deliver and execute an effective security strategy that actually deals
with
problems rather than sweeping them under the carpet or waste time by
blaming
others for their own failings. Oracle's CSO has had five years to make
improvements to the security of their products and their security
response
but in this time I have seen none. It is my belief that the CSO has
categorically failed. Oracle security has stagnated under her
leadership and
it's time for change.
I urge Oracle customers to get on the phone, send a email, demand a
better
security response; demand to see an improvement in quality. It's
important
that Oracle get it right. Our national security depends on it; our
companies
depend on it; and we all, as individuals depend on it.
Cheers,
David Litchfield
Intro
