Slashdot Mirror
Microsoft: As of October, 1024-Bit Certs Are the New Minimum
way2trivial writes with this snippet from Information Week about a warning from Microsoft reminding Windows administrators that an update scheduled for October 9th will require a higher standard for digital certificates. "That warning comes as Microsoft prepares to release an automatic security update for Windows on Oct. 9, 2012, that will make longer key lengths mandatory for all digital certificates that touch Windows systems. ... Internet Explorer won't be able to access any website secured using an RSA digital certificate with a key length of less than 1,024 bits. ActiveX controls might be blocked, users might not be able to install applications, and Outlook 2010 won't be able to encrypt or digitally sign emails, or communicate with an Exchange server for SSL/TLS communications."
Thank you for being a friend
Traveled down the road and back again
Your heart is true, you're a pal and a cosmonaut.
And if you threw a party
Invited everyone you knew
You would see the biggest gift would be from me
And the card attached would say, thank you for being a friend.
First!
If MS are really interested in everybody's security why don't they close the Windows Update backdoor instead?
System have the ability to go further, why not make 2048 the minimum? Does anyone know why 1024 was selected? I would guess it has to do with some backwards compatibility with something. Some of the issuers are making it next to impossible to go below 2048.
I don't really understand how anyone can care whether a closed source operating system is secure.
No matter how few people actually read through the Linux kernel code, it's sufficiently open that blatant backdoors are not going to be inserted. Why would I assume the same of Windows or (yet worse, because they have some quasi-religious thing going on) OS X?
Way to go Microsoft! As everyone moves to 2048 bit keys, let me be the first to welcome you to 2002.
We could've gotten notification a week or two after the update.
This is gonna be a pain in the butt though =\
On the bright side things will be more secure!
TechRepublic noted this a while ago and provided detailed instructions on how to work-around the issue.
"Maybe this world is another planet's hell"
Aldous Huxley
Wouldn't be much of an OS if it didn't have a reach-around.
The rest of the industry is already moving to 2048 bit. Is it really so hard for MS to get their heads out of the sand and actually look at what's going on outside of Redmond for a change? It's ridiculous. Like the tail wagging the dog. Practically every single computing niche outside of the desktop is dominated by some form of Unix or Unix-like OS. From supercomputers, to cellphones and everything in between. When is MS going to stop being a drag on this industry? Damn, as much as Apple sucks, I think the world will be better off if everybody just stops using desktops and goes to iPads. A lot more secure too!
Just switch to Linux. Do you really need them telling you what you have to do?
* Carthago Delenda Est *
That many institutions have access to MS Source Code is kinda like instituting a needle-inna-haystack search.
Yes you might find a needle, but unless you're a needle-collector or perhaps a seamstress what in this universe d'you think you're gonna do with it?
At least with Open Source you can
(1) fix the problem with the code
(2) submit the code back to The Author
(3) expect that The Author will either accept the fix as is or perhaps integrate the solution with more elegance
Sure not *always* but the expectation would be more-often-than-not your fix (in one form or another) reaches the wider community of users.
Visit CryptoGnome in his home.
No matter how few people actually read through the Linux kernel code, it's sufficiently open that blatant backdoors are not going to be inserted.
Open source suffers from quasi-religious stuff too, as you just demonstrated with your claim. Ken Thompson, of Bell Labs and Unix and C fame - the "K" in K&R, demonstrates the insufficiency of being able to read the source code.
http://cm.bell-labs.com/who/ken/trust.html
The reason Thompson's backdoor is famous is because it was far from blatant.
Sounds like a great way to get people off IE, or fill up customer service inboxes for weeks. Madness!
Really? Wow. Way to go and show your intellect.
But isn't that the point? Just imagine what you can hide in something as large as the Linux kernel or most modern programs.
There are multiple Microsoft keys with a size under 1024-bit out there in the wild, and certificate chains involving them they were used in state-sponsored attacks. It is therefore quite correct, and very necessary, of them to reject RSA keys with a crackable length. Keys up to 768 bits have been cracked publically.
What I'm mystified about is that 1024-bit RSA keys are still allowed as a baseline - honestly, those should all be phased out already and I haven't considered them safe enough for over half a decade now. While no-one has publically factored the RSA-1024 test vector, estimates in 2007 showed that it would indeed be possible, and tests proved positive - 5 years on, I expect it to be quite feasible to factor a 1024-bit RSA key now, particularly if you implemented parallel parts of the sieve in GPU shaders and used something like... oh, I don't know... the NSA's new Tesla-based supercomputer. And they're far from the only ones: Iran has either factored or swiped at least one 1024-bit key (honestly, either is plausible at this point).
This is why 2048-bit keys going onwards are all that's allowed by CAs (and were mandated from the beginning of the EV standard). In practice this has never been a problem - they'll work on any version of Windows which supports RSA at all. I remember using 4096-bit keys with Windows 95, and indeed I recall experimental builds of PGP happily using 16384-bit RSA keys (although they were, of course, slow as molasses).
Long-term you should probably think about moving to prime256v1 (secp256r1). That's got more juice than 3072-bit RSA, but it's vastly faster and much smaller. Alternatively, curve25519/ed25519, which are extraordinarily optimised binary curve algorithms with nice features such as being secure through the twist, and not needing a random source to create signatures.
No one can factor my 1024-bit cert: 0x0000.....000F.
Open source is great mechanism for finding security holes, but it's hardly the only mechanism. OK, Windows is probably not as secure as Linux, but it's not totally insecure either.
Hey, I live in an apartment that doesn't have the best security, but enough for the neighborhood in which I live. By your logic, I should either beef up security to the max (iron bars on the windows, install a CCTV, maybe get a pit bull) or just forget all about it and leave never lock the front door or window by the fire escape. Makes no sense
Wake up and smell the Firefox/Safari/Chrome grandma!
The "K" of K&R is wrong.
"K" is Brian Kernighan. You know, the Brian Kernighan of "The C Programming Language" fame. He wrote a book or two. He's quite famous. Maybe you've heard of him.
Look it up.
Grandpa: My Homer is not a communist. He may be a liar, a pig, an idiot, a communist, but he is not a porn star.
There is an entire collection of root certs in your browser that are all trusted unconditionally. Hundreds of them, in fact. These root certs have signed thousands (who knows how many, really?) intermediate certs. All of these intermediate certs are trusted unconditionally to authenticate any SSL server whatsoever. It's pointless to have a key longer than the shortest intermediate cert key length in use anywhere. When you use SSL, you are trusting thousands of unknown parties with absolute cert-signing authority. SSL certificates are known to have been used for explicit man-in-the-middle purposes: Trustwave sold root certificate for surveillance. Sure they revoked that one key because of the bad publicity, but it's common industry practice. How is SSL hopelessly broken? Let us count the ways.
There's a reason why most coding standards explicitly disallow assignment inside a conditional structure.
I've never heard of this. Are you actually trying to say that most coding standards disallow if(condition) { var = true; } else { var = false; }? Or do you actually mean variable declarations when you say assignments? Could you link to some coding standard that actually disallows this, so I could look into this a bit more?
Your post has nothing at all to do with mine apart from some sort of personal attack based on the assumption of extreme ignorance ("It might suprise you" followed by the fucking incredibly obvious) and some fallacious assumption that the above is in some way an attack on others ("that programmers at Microsoft are all incompetent is partisanship of the most extreme kind" - WTF? How do you find an accusation of incompetance in the two sentence post above?). Sorry limecat - it's coming off as insane as saying I must hate tigers because I like dogs. If you want to argue with voices in your head please do not commit the gross insult of pretending that I am one of them.
Up to a point fragmentation or variety is a good thing. And not just in software. In agriculture, if your field consists of only one crop, your goose is cooked if there's an outbreak of a plant disease. A country whose GDP comes from a single source, say oil or a single cash crop, is also more vulnerable to price fluctuations in the global market. A crash in the prices of that product would lead to a crash in the country's economy as well.
Too much fragmentation of course is bad. But as far as Linux, the major distros are quite few, namely, Ubuntu, Redhat, Fedora, Debian, and possibly Suse. It's their derivatives that give the impression of excessive fragmentation. Derivatives tend to be compatible with the mother distro at least as far as the installation of third party programs not in the main repository. A binary-only printer driver that can run in Ubuntu would be compatible with Linux Mint for example.
If you use a program like SCCM, SCE, EmminantWare/SolarWinds, Secunia, Local Update Publisher (plug: my OSS alternative), or any other similar program that allows you to publish your own packages through the WSUS system you will also need to worry about this. For some time the default certificate that gets created was 512 bits and will become invalid with this update. Check with your vendor to see what remedy they suggest. One of the recent updates to the WSUS API bumped this default cert creation to 2048 bits but that won't help existing users.
I'm the head IT manager for this 50-person company so I'm stuck as the server administrator despite having about 12 weeks of MCSE training. Someone else set up our current self-signed certificate so I don't know the size or how to check. I do know they plopped the .cer file itself down on C: though, lol. So I opened it in notepad, pasted its main contents to Word, and ran a character count. It's 2092 total characters in size. I'm going to take a guess that that's a 2048 bit cert, right? So: :-P
1. Am I correct in assuming that?
2. Is there a less stupid way of checking that in Server 03?
3. I have no idea where our other server (exchange 03) keeps its cert or what size it would be expected to be by default. Anyone got some info on that?
Disallowing assignments in conditions would work too.
A lot of C compilers already have compile-time options to warn when the outermost operator in an if or while condition is an assignment. This allows various idioms where conditions have side effects, such as pulling one pointer from an iterator function and then checking whether or not it's a null pointer terminator, while requiring the programmer to make his intent explicit.
Their hiring process showed me very clearly that they had no interest in my creativity and only wanted someone who can churn code, find bugs, and patch systems to keep them running (all important; but not the only thing in the world; and definitely not for me).
I believe it's called "paying your dues", and I believe it's considered standard practice at most major companies.
With closed source you can leverage the SLA's between yourself and the vendor
With closed source, there is only one provider of such SLAs for each product. With open source, you can leverage the competition among several companies that offer SLAs for a given open source platform. For example, one can choose Red Hat, Oracle, or Canonical to support a Linux deployment.
Eat that!
I'm not sure a kilobyte-strength Cert would be strong enough to freshen the breath of some of the people I've worked with...
According to GlobalSign (one of the largest CA's), they stopped issuing 1024bit keys back in 2010... The lowest encryption they (and most CA's) use now is 2048bits. https://www.globalsign.com/support/faq/sslfaq.php All orders placed from November 29th 2010 will only be accepted with a CSR key length of 2048 bits or higher. This is to fully comply with the National Institute of Standards and Technology Recommendations (NIST) and the mandatory requirements by Microsoft's Root Certificate Program to issue Certificates from a minimum of 2048 bits by January 1, 2011. Maybe it's related to the Y1969 bug :)
The day Microsoft creates a product that doesn't suck, it will be known as the Microsoft Vaccuum Cleaner!