The stories center around how open source software such as Firefox was able to produce updates to correct this issue just a few days after the Blackhat conference, while Microsoft still hasn't fixed it and are "investigating a possible vulnerability in Windows presented during Black Hat".
But the actual timeline is missing from these stories.
The NULL character certificate flaw (CVE-2009-2408) was actually disclosed by two researchers working independantly who both happened to present the work at the same conference, Blackhat, in July this year. Dan Kaminsky mentioned it as part of a series of PKI flaws he disclosed. Marlinspike had found the same flaw, but was able to demonstrate it in practice by managing to get a trusted Certificate Authority to sign such a malicious certificate.
The flaw was no Blackhat surprise; Dan Kaminsky actually found this issue many months ago and responsibly reported the issues to vendors including Red Hat, Microsoft, and Mozilla. We found out about this issue on 25th February 2009 and worked with Dan and some of the upstream projects on these issues in advance, so we had plenty of time to prepare updates and this is why we were able to have them ready to release just after the disclosure.
We keep all our friends and family contacts in a single text file in vCard format. We sync this file to our phones (mobile and house DECT phones) and home automation system (for caller ID and phone book). I also print out a copy to take when travelling. Except I rarely print out an update as I've failed to find any useful program to pretty print the contacts. Previously I used a quick hack script in perl to convert the vcard entries to HTML, but it wasn't clever enough to handle page breaks and needed manual setting all the margins and page sizes correctly. I like to print it to fit in my paper planner, a Compact size Franklin Covey planner system.
I've been using Scribus for a few months, mostly for our wedding invites and stationary, and spotted that Scribus had a Python API. So a few hours later and out has popped a Python script you can use to pretty print a vCard vcf file, handling page breaks, images, and large margins to skip the hole punches.
Here is an extract from a sample vCard file:
BEGIN:VCARD ADR;TYPE=work:;;10 Downing Street;London;SW1A 2AA TEL;TYPE=fax:+44 2079 250918 NICKNAME:Prime Minister FN:Gordon Brown N:Brown;Gordon PHOTO;VALUE=URI:http://www.number10.gov.uk/wp-content/uploads/pm-official-pic-234x300.jpg VERSION:3.0 END:VCARD
You'll need a few things:
Use the "Script" "Execute Script" option, find and select vcf2scribus.py and hopefully you'll end up with something like this:
You can then save it as a pdf or print it direct.
The script is a bit of a hack and has hard-coded page sizes, fonts, margins, vcard sections used, and so on. But I figure it might save someone a couple of hours and only needs a bit of modification to suit. It would be fairly easy to extend the script to use the Scribus API to let folks select the vcard file, page sizes, fonts, and things. Bonus points if you fix it to figure out the final sizes of the images and right align them. This is my second ever python program so no sniggering at the code!
I met Tracy via ICQ over nine years ago and we've been partners for eight, so it was about time we got around to getting married. We both wanted a castle wedding, and as Monty Python fans we couldn't resist Doune Castle, a restored 14th century stronghold not far from Glasgow which was used in the "Holy Grail" film.
So we couldn't resist a few Python moments, with coconuts, the French guards door scene, and the killer rabbit of Caerbannog. Run away!
Our cake topper was commissioned from an artist in Brazil we found via flickr.
All the artwork including invitations, day plans, and place cards was drawn and created by ourselves using Inkscape and Scribus on Fedora, I'm going to blog about that and share the files when I have some more time next month.
The only issues we had on the day where with the cake which the supplier (Marks and Spencers) lost part of and were unable to replace with the right cake, and a poor substitute DJ (our chosen DJ ended up in hospital just before the event).
Everything else went amazingly well and to plan and despite the poor August weather in Scotland this year managed great weather with only a light shower as guests were leaving the castle.
The chart below illustrates the total number of security updates issued for Red Hat Enterprise Linux 5 Server as if you installed 5.3, up to and including the 5.4 release, broken down by severity. I've split it into two columns, one for the packages you'd get if you did a default install, and the other if you installed every single package (which is unlikely as it would involve a bit of manual effort to select every one). For a given installation, the number of package updates and vulnerabilities that affected you will depend on exactly what you have installed or removed.
So for a default install, from release of 5.3 up to and including 5.4, we shipped 51 advisories to address 166 vulnerabilities. 8 advisories were rated critical, 18 were important, and the remaining 25 were moderate and low.
Or, for all packages, from release of 5.3 to and including 5.4, we shipped 78 advisories to address 251 vulnerabilities. 9 advisories were rated critical, 28 were important, and the remaining 41 were moderate and low.
The 9 critical advisories were for just 3 different packages. In all the cases below, given the nature of the flaws, ExecShield protections in RHEL5 should make exploiting these memory flaws harder.
Updates to correct all of these critical vulnerabilities were available via Red Hat Network either the same day, or up to one calendar day after the issues were public.
In fact for Red Hat Enterprise Linux 5 since release and to date, every critical vulnerability has had an update available to address it available from the Red Hat Network either the same day or the next calendar day after the issue was public.
Although not in the definition of critical severity, also of interest during this period were several NULL pointer dereference kernel issues. NULL pointer dereference flaws in the Linux kernel can often be easily abused by a local unprivileged user to gain root privileges through the mapping of low memory pages and crafting them to contain valid malicious instructions:
Red Hat Enterprise Linux since 5.2 has contained backported patches from the upstream Linux kernel to add the ability to restrict unprivileged mapping of low memory, designed to mitigate NULL pointer dereference flaws. However it was found that this protection was not sufficient, as a system with SELinux enabled is more permissive in allowing local users in the unconfined_t domain to map low memory areas even if the mmap_min_addr restriction is enabled. This is CVE-2009-2695 and will be addressed in a future kernel update.
Red Hat Enterprise Linux 5 shipped with a number of security technologies designed to make it harder to exploit vulnerabilities and in some cases block exploits for certain flaw types completely. From 5.3 to 5.4 there were three flaws blocked that would otherwise have required critical updates:
To compare these statistics with previous update releases we need to take into account that the time between each update is different. So looking at a default installation and calculating the number of advisories per month gives the results illustrated by the following chart:
This data is interesting to get a feel for the risk of running Enterprise Linux 5 Server, but isn't really useful for comparisons with other versions, distributions, or operating systems -- for example, a default install of Red Hat Enterprise Linux 4AS did not include Firefox, but 5 Server does. You can use our public security measurement data and tools, and run your own custom metrics for any given Red Hat product, package set, timescales, and severity range of interest.
The first issue is that many web browsers still accept certificates with MD2 hash signatures, even though MD2 is no longer considered a cryptographically strong algorithm. This could make it easier for an attacker to create a malicious certificate that would be treated as trusted by a browser. It turns out that there are not many valid MD2 hash certificates around any more, and the main one that does exist is at the trusted root level anyway (and there is actually no need for a crypto library to verify the self-signature on a trusted root). So most vendors have chosen to address this issue by disabling MD2 completely for certificate verification. This is allocated CVE name CVE-2009-2409 ( single name for all affected products).
OpenSSL. For upstream OpenSSL we have disabled MD2 support completely. This was done in two stages; the first was a patch in June 2009 that removed the redundant check of a trusted root self-signed certificate. Then in July, MD2 was totally disabled. So this issue does not affect OpenSSL 1.0.0 beta 3 or later. Although there have not yet been an upstream release of 0.9.8 containing this fix, a future OpenSSL 0.9.8 (after 0.9.8k) will disable MD2, probably in a few weeks.
GnuTLS. The upstream GnuTLS library has for some time meant to have disabled MD2 support, although due to a broken patch it wasn't actually disabled correctly until January 2009. So this issue does not affect GnuTLS versions 2.6.4 and above, or GnuTLS versions 2.7.4 and above.
NSS (and hence Firefox). The upstream NSS library since version 3.12.3 (April 2009) has disabled MD2 and MD4 by default (although legacy applications could turn it back on using an environment variable "NSS_ALLOW_WEAK_SIGNATURE_ALG" if they need to). Mozilla Firefox since version 3.5 has used this NSS version and therefore MD2 is disabled. I suspect this issue will get addressed in a future Firefox 3.0 update in the future too if they rebase to the new NSS.
There is no immediate panic to address this issue as a critical security issue, as in order for it to be exploited an attacker still has to create a MD2 collision with this root certificate; something that is as of today still a significant amount of effort.
My CVSS v2 base score for CVE-2009-2409 would be 2.6 (AV:N/AC:H/Au:N/C:N/I:P/A:N)
This issue is about how Common Names are checked for validity by applications. For example if a server presents a certificate with two CN entries, how does the app validate those. Does it use the first one, the last one, or all of them?
OpenSSL. OpenSSL provides an API that allow applications to check CN names any way they want. It turns out that, without sound guidance, applications have tended to do things differently. A summary of a few OpenSSL applications is in this Red Hat bugzilla comment. But as a CA should validate all CN names in a certificate being signing, these are really just bugs and do not have a security impact
The second issue is all about inconsistencies in the interpretation of subject x509 names in certificates. Specifically "issue 2b, subattack 1" is where a malicious certificate can contain leading 0's in the OID. The idea is that an attacker could add in some OID into a certificate that, when handled by the Certificate Authority, would appear to be some extension and ignored, but when handled by OpenSSL would appear to be the Common Name OID. So the attacker would present the certificate to a client application and it might think that the OID is actually a Common Name, and accept the certificate where it otherwise should not.
OpenSSL. This is not a security issue for OpenSSL. Steve Henson explains: "OpenSSL does tolerate leading 0x80 but it does _not_ recognize this as commonName because the NID code checks for a precise match with the encoding. Attempts to print this out will never show commonName nor will attempts to look up using NID_commonName". However this will be addressed as a bug fix in the future.
NSS (and hence Firefox). NSS is noted in the paper as having a similar issue, but again it's not fooled into treating the OID as a Common Name so this is not a security issue (and therefore I didn't check if this is already fixed in the new upstream NSS).
"issue 2b, subattack 2" is where a malicious certificate can have a very large integer in the OID. The idea is that an attacker could add in some OID into a certificate that, when handled by the CA, would appear to be some extension and ignored, but when handled by OpenSSL would overflow and appear to be the Common Name OID. So the attacker would present the certificate to a client application using OpenSSL and it might think that the OID is actually a Common Name, and accept the certificate where it otherwise should not.
OpenSSL. This issue was actually fixed upstream in September 2006 in OpenSSL 0.9.8d by switching to using the bignum library for handling the OID. Even for older versions though it's really not a security issue for the same reason as given earlier: the OpenSSL NID code checks for a precise match with the encoding. So attempts to print this out will never show it being a Common Name, nor will attempts to look it up as a Common Name succeed.
"issue 2, attack 2c" is regarding NULL terminators in a Common Name field. If an attacker is able to get a carefully-crafted certificate signed by a Certificate Authority trusted by a browser, the attacker could use the certificate during a man-in-the-middle attack and potentially confuse the browser into accepting it by mistake.
NSS (and hence Firefox). This issue affected NSS and is assigned CVE-2009-2408. The upstream NSS library since version 3.12.3 (April 2009) has fixes to address this issue. Therefore Firefox 3.5 is not affected.
My CVSS v2 base score for CVE-2009-2408 would be 4.3 (AV:N/AC:M/Au:N/C:N/I:P/A:N)
"issue 2d" is how the OpenSSL command line utility will output unescaped subject X509 lines to standard output. So if some utility runs the openssl application from the command line and parses the text output, and if an attacker can craft a malicious certificate in such a way they fool a CA into signing it, they could present it to the utility and possibly fool that utility into thinking fields were different to what they actually are, perhaps allowing the certificate to be accepted as legitimate.
OpenSSL. This attack requires that some utility will parse the output of OpenSSL command line using the default 'compat' mode. Applications should never do this. Upstream OpenSSL are unlikely to address this issue directly, although in the future the default output mode perhaps could be changed to something other than 'compat', and it's likely a documentation update will remind users that parsing the output of running such an openssl command is not the right way to use OpenSSL.
OpenSSL. This was reported by Dan to OpenSSL and fixed in March 2009. This was allocated CVE name CVE-2009-0590. Therefore OpenSSL 0.9.8k and later contains a fix for this issue.
My CVSS v2 base score for CVE-2009-0590 would be 2.6 (AV:N/AC:H/Au:N/C:N/I:N/A:P)
Last winter the Red Hat video team came to my house to record some footage for both internal and external use. On one of the internal videos they look at my home automation system, point the camera at a wall tablet, and figure out that I'd not used my treadmill in over two years. So there were really two options (1) remove the year from the display so it would never look like we were slacking for more than a year, or (2) find a way to get motivated again.
Recently we both started using Twitter, so it seemed like a natural progression to hook the treadmill to twitter and have it publicly embarrass us for slacking off.
We called it 'twedmill' ('tweadmill' perhaps is more correct, but just sounds like a factory that weaves twead jackets). Here is how it works:
The treadmill itself is pretty standard; it's from Trimline and has a fancy
computer. When I looked inside and saw a PIC I was tempted to interface direct
to the computer, but didn't really have the time to get around to that.
Although the treadmill does things like have a variable incline and measurement of heart
rate, all I really care about it making sure we were using it, for how long,
and how far we got.
Under a cover in the base are the PWM controllers, motors, and the belt drive to the treadmill deck. The treadmill itself measures the belt speed by having a single magnet on the wheel and a small sensor next to it, one revolution giving one pulse. So to keep things simple I just hot-glued a spare reed switch I had around so the same magnet would trigger it. The reed switch happily copes with the treadmill even on top speed, so no real need for anything more fancy.
I didn't have anything that could accurately measure the diameter of the roller, so
by counting pulses at various speeds and comparing to the onboard
display it worked out at 8122 pulses/revolutions per (uk) mile (so that's
about 198mm of travel per pulse, making the diameter of the
roller about 63mm).
I use a 1-wire network in the house to measure temperatures, watch the doorbell, and control the central heating system, so I wanted to use the same system to deal with the treadmill. So the reed switch connects to a DS2423 counter (Unfortunately it seems the DS2423 is discontinued now). The DS2423 was only available in a surface-mount package, so I found some converters on ebay to save having to design a PCB just for three components. The DS2423 connects into a 1-wire hub in node0, then to a 1-wire USB adapter on our main server, currently running Fedora 10.
The software used in based on the source code from 'digitemp'
as it includes
cnt1d.c to read the counter values. Every ten
seconds the jabber treadmill bot switches to the right network segment
on the 1-wire hub then polls the counter of the DS2423 to see
if the treadmill has moved. Once the treadmill has stopped moving for
a while the software stores the total distance travelled and time in
a database, sends an email, and uses the perl Net::Twitter module to
post a mesage to twitter. (It can also draw a graph showing speed over
time, but that turned out to be not very interesting)
For the future I'd quite like to hook directly into the treadmill computer, perhaps giving two way control of the treadmill programs, as well as recording the incline and heart rate. Another idea has been to use the current treadmill speed to decide which music video to play next based on bpm (the tv is connected to an old XBOX running XMBC so could easilly be remotely controlled to switch videos). Or perhaps link it to google streets for a virtual jog through some random town. Finally, you currently have to select who is using the treadmill before (or very quickly after) using it using the touch panels in the house; which seems like a good excuse to play with some RFID in our shoes, perhaps also using that to select a playlist of music videos per person.
Having a CCV of 000 seems nice and easy to remember, but actually was a bit of a curse. To start with, companies would sometimes not believe that 000 is your real CCV when you tell them by phone. But usually after a few attempts you can convince them to at least try it, and then all is well.
The real problems came when using the card online as several merchants refused to accept the card. Any programmer reading this will have guessed the ways this could fail already. Rather than web applications checking for a CCV of three digits, I imagine some of them stored the field as an integer and had "0" overloaded as "didn't enter a CCV".
Scan Computers was the first casualty; my first order with them using the card appeared to get accepted, but then got stuck and the order stalled. That took a phone call to sort out, but at least the guy I spoke to by phone recognised and understood the problem and I only ended up getting my stuff a day late. It's worked okay with them since, I guess they fixed it.
Some other merchants I've been less lucky with. Some refused to accept the CCV at the time I entered it, but at least with those you know immediately and can use a different card. Other merchants accepted the CCV at the order time but then later rejected the order usually without giving a reason; probably when they did some batch processing with the stored CCV.
So you'd think there would be a lot of people with this problem: if the CCV is generated by the issuer using some hash then it ought to be 1/1000th of the card holding population. Perhaps some issuers deliberately avoid giving out a 000 security code, or perhaps I was just unlucky in my choice of merchants.
The experiment has sadly come to an end now as the card expired and was been replaced by one with a different CCV. I'm hoping one day to get 999.
From time to time I publish metrics on vulnerabilities that affect Red Hat Enterprise Linux. One of the more interesting metrics looks at how far in advance we know about the vulnerabilities we fix, and from where we get that information. This post is abstracted from the upcoming "4 years of Enterprise Linux 4" risk report
For every fixed vulnerability across every package and every severity in Enterprise Linux 4 AS in the first 4 years of its life, we determined if the flaw was something we knew about a day or more in advance of it being publicly disclosed, and how we found out about the flaw.
For vulnerabilities which are already public when we first hear about them we still track the source as it's a useful internal indicator on where the security response team should focus their efforts.
So from this data, Red Hat knew about 51% of the security vulnerabilities that we fixed at least a day in advance of them being publicly disclosed. For those issues, the average notice was 21 calendar days, although the median was much lower, with half the private issues having advance notice of 9 days or less.
The chart below shows the total number of security updates issued for Red Hat Enterprise Linux 5 Server as if you installed 5.2, up to and including the 5.3 release, broken down by severity. I've split it into two columns, one for the packages you'd get if you did a default install, and the other if you installed every single package (which is unlikely as it would involve a bit of manual effort to select every one). So, for a given installation, the number of packages and vulnerabilities will probably be somewhere between the two.
So for a default install, from release of 5.2 up to and including 5.3, we shipped 45 advisories to address 127 vulnerabilities. 7 advisories were rated critical, 21 were important, and the remaining 17 were moderate and low.
For all packages, from release of 5.2 to and including 5.3, we shipped 61 advisories
to address 181 vulnerabilities. 7 advisories were rated critical, 28 were
important, and the remaining 26 were moderate and low.
The 7 critical advisories were for just 3 different packages:
Although not of critical severity, also of interest during this period were the spoofing attacks on DNS servers. We provided an update to BIND (July) adding source port randomization to help mitigate these attacks.
Updates to correct all of these critical vulnerabilities (as well as migitate the BIND issue) were available via Red Hat Network either the same day, or one calendar day after the issues were public.
In fact for Red Hat Enterprise Linux 5 since release and to date, every critical vulnerability has had an update available to address it available from the Red Hat Network either the same day or the next calendar day after the issue was public.
To compare this with the last updates we need to take into account that the time between each update is different. So looking at a default installation and calculating the number of advisories per month gives the following chart:
Red Hat Enterprise Linux 5 shipped with a number of security technologies designed to make it harder to exploit vulnerabilities and in some cases block exploits for certain flaw types completely. For 5.2 to 5.3 there were two flaws blocked that would otherwise have required updates:
This data is interesting to get a feel for the risk of running Enterprise Linux 5 Server, but isn't really useful for comparisons with other versions, distributions, or operating systems -- for example, a default install of Red Hat Enterprise Linux 4AS did not include Firefox, but 5 Server does. You can use our public security measurement data and tools, and run your own custom metrics for any given Red Hat product, package set, timescales, and severity range of interest.See also:5.1 to 5.2 risk report
Actually this isn't surprising and is exactly what I'd expect; it's all down to third party applications.
Let's say you're browsing the web. It's more than likely that at some point you'll want to view some PDF files, watch some Flash content, or play a Java game. Those tasks are all dealt with by third party applications, although to the end user it's all part of the browser experience. Since your system is only as secure as its weakest link, you need to manage security updates for those third party applications just as carefully as you manage security updates for the rest of your system. That's why Adobe Reader, Java, Flash, and all the myriad of other applications you've installed in order to make your system useful have their own update mechanisms. Some applications on Windows will 'phone home' when they are run and check to see if they need to be updated, others deploy services that sit in the background looking for updates from time to time, others even check every time your system starts. Many don't get automated updates at all.
How do you deal with all that risk? I believe it's possible by providing an OS distribution which includes all the bits you'll likely need to make a useful computing environment, thereby taking away that update uncertainty. Red Hat ship several PDF viewers in our distributions for example, but we also ship (in an Extras channel) Adobe Reader. Our Security Response Team are monitoring for security issues in everything we ship, all the third party applications, and providing a single point of contact, a single notification system, and a single way to get the updates.
If Microsoft knew that say 25% of all their users installed Firefox, wouldn't they be better bundling it and providing their centralised automated updates for it, to reduce their customers overall risk? They do already bundle some third party applications, although it's been with mixed success as we found 3 years ago when they didn't provide security fixes for bundled Flash (ZDNet coverage).
This is, in part, why you've not seen me respond recently to the Vista security reports which compare vulnerability counts. In these reports they use a cut-down minimal Red Hat Enterprise Linux installation in order to make it look more like Windows for the comparisons. But this is completely backwards -- the fact that we're including and fixing the flaws using a common process in so much third party software is actually helping reduce the risk and protect real customers. For example we could easily cut our vulnerability count by shipping only one PDF viewer instead of four. But if we know that these other viewers are going to get installed by the customer anyway all we've done is to hide the vulnerability count elsewhere, and you've made the customers overall risk increase.
So it may seem counter-intuitive but we should ship as much third party applications (that we know people use) as we can, because a single managed security update and notification process will decrease a users overall risk. The fewer third party applications that users have to get from elsewhere and install and manage for themselves the better in my opinion.