Slashdot Mirror
The Death Throes of crypt()
dex writes "Tom Perrine and Devin Kowatch of the San Diego Supercomputer Center have issued "Teracrack:
Password cracking using TeraFLOP and PetaByte Resources" (PDF, HTML version via Google). Using SDSC's
prodigious computing facilities, they precomputed 207 billion crypt() hashes in
80 minutes."
I wonder if this will spur Sunto finally make the default password encryption algorithm on Solaris something other than crypt...
Actually with most Unixish systems going to other password formats such as MD5 and Blowfish I'd think that this goes to show that (NSA notwithstanding) crypt() has had a long, healthy existance. Rather than saying 'crypt() is dead' they should be saying 'it took 30ish years but crypt() is at the end of its useful life'.
Not many pieces of code will be able to boast that lifespan.
Trolling is a art,
Unless they release these hashes out into the wild, the average cracker/hacker does not have access to this type of resource...
Definately cool though for proof of concept!
80 Minutes? Obviously we just are not using enough power.
30% Troll, 50% Underrated, 10% Interesting
Score:5, Troll
Your name is ObviousGuy for a reason then, huh?
Occam's razor is the blind faith in the natural selection of least resistance and in universal oversimplification. -- EF
In the wake of stories like this, and yesterday's story about RSA-576 being cracked (here), is this a message that we need more secure forms of encryption than we already have? RSA is great so far, but how long until 1024 is broken? Or any other schemes, like the MD5 hashing that's used for digital signatures?
Topics that people with lots of credentials behind their names are going to have to solve. Keeping ahead of the crackers is a big concern not only for security of transactions, but for personal privacy as well.
Erioll
Mirrored here:
http://home.twcny.rr.com/scooper2/teracrack.pdf
- Sometimes you're the pidgeon, sometimes you're the statue.
This should cause some interesting systems integration issues as crypt has become the defacto standard for cross system authentication and password management. (hash it at your web server, compare it with app server, store it in DB, where it is used by samba to auth winblow users, blah blah, I know these arent exact implementation examples but you get the idea). Just a lot of code or libraries to change to make a system secure.
Remember that many glibc2 "overloads" crypt() to be able to deal with MD5 hashes. I wonder how many MD5 hashes they can generate...
That over time, any encryption alghorythm may be broken by superior computer. 50 years from now, normal computers will put anything we have to shame, and supercomputers will make current ones look like calculators.
Crypt is already supplantable by many improved techniques, but even if it is used, are they going to make these keys available to the world?
If not, now that it's known a really faster computer can solve then, perhaps the next step in spammy-crackers' arsenal will be to take their virussed drones away from attacking anti-spam sites and focus them at generating crypt or other password solutions? How many drones working P2P-style (you create these hashes, I'll create these ones) would it take to equal this supercomputer?
I think the answer to that question is obvious, guy.
Clearly, crypt() was meant to die: just look at its name!
As Schneier says on the first page of Chapter 1 of "Applied Cryptography",
They've got the tables on their ftp server, but it seems slashdotted because it's going really slow... my computer says "downloaded 4194304 bytes of 1209462790550 bytes (0.00034%)"
Anyone have a bit torrent for this thing?
HIV Crosses Species Barrier... into Muppets
If I understand the article correctly, they're using serious computer power to develop a database of all passwords and their resulting hashes. In doing so, they can reverse engineer any hash back into a password via a 1 to 1 lookup. The only problem is that the attacker still needs physical access to your password hashes. In other words, this is much more serious for "insider" hacks where a system user wants root control.
Correct me if I'm wrong, but wasn't the point of moving the hashes from passwd to another root protected file (like "master.passwd") to prevent this sort of problem?
Javascript + Nintendo DSi = DSiCade
The ability to generate lots of crypt strings only helps you if you have the original crypt string to compare against. Most modern UNIX systems store crypt strings in /etc/shadow which is only readable by root. The crypt string is never passed across the net during most auth sequences. (Certain types of LDAP auth being the exception here.)
The problem occurs if someone manages to break into a machine, achieve root, and pick up the /etc/shadow file. They can now brute-force all the passwords given enough time, and it appears that the amount of time needed is shrinking.
This is a good argument for using different passwords on untrusted boxese and changing your password often.
Even so, using a 10 character input of about 72 possible input chars, isn't 207 billion still only like .0000055% of the total search space?
So that 20000 * 80minutes gives ~1% of the space cracked?
2000000 * 80 minutes = 304 years to fully close the space.
With a perfect distribution, the mean of about 150 years seems like a long time.
Someone please check my assumptions here.
Sure, MD5 strings are a little longer, but still (I'm pretty sure) less than 80 characters.
md5 strings are 32 characters. And most linux distros use md5 for their shadow files already.
For a book called Applied Cryptography. Perhaps he should have called it Applied Cipherology. Applied Ciphering? Applied Cipherography? I think ill just stick with "crypt", thanks.
I remember back in high school I had created a bunch of accounts on my linux box and used some program to try and decipher the passwords. 1-4 charecter passwords were found in 30 minutes (on my blazing-fast-at-the-time 200MHz Pentium 1), 5 charecter passwords took 2 days, 6 charecter passwords took... well, forever more or less. I figured at the time, a 7 charecter password would be sufficient forever (at least for my life time), but I guess not. Now I use 10 charecter passwords for most of my stuff... Do I need to move to 15 charecters? A passphrase instead of a password?
The Doormat
If you're not outraged, then you're not paying attention.
Wait a year or three and this kind of computing power will be available in game consoles in bedrooms across america.
I've already rooted all your boxen and converted them to a worldwide Beowulf cluster.
Time to crack some pr0n passwords...
Heheh... the paper actually talks about them putting a searchable front-end to the results online but then says they decided not to, in part due to the "dreaded 'slash-dot' effect". Nice.
Moral of the story: Pick a good password.
Funny, we did this work almost a year ago, and someone finally notices :-(
:-)
Yes, this is a very small part of the total theorectical key space.
But its exactly the part of the space that is most "interesting"; this is the part that will most likely be searched by an attacker AND the part that is most likely to have a real, user-selected password.
The original goal (years ago) was to allow us to verify that our users weren't using passwords that would be likely to be found by an attacker.
Yes, passwords are *supposed* to be stored in shadow files that are not accessible, except by root, but in practice, it is often discovered out in the Real World, especially at larger multi-vendor sites, that user password hashes are copied between machines, stored in databases, and in general available to an attacker who does not yet have root.
These are the same sites that often can't or haven't converted to md5() hashes, as they have older legacy OSes that don't do md5(). Note that even though Sun "supports" md5() hashes, they don't support them everywhere and it certainly isn't seamless. Don't get me started about AIX. Linux and the *BSD folks are way ahead of most of the commercial UNIX variants.
As for the scalability, read the whole paper. We used the largest single machine at SDSC, but its rather dated in terms of crypt() performance. A distributed.net-style project using typical home machines would win, IF you could get a thousand or so people to cooperate.
The Terabytes of storage in a single filesystem didn't hurt in the sort/merge phases either.
Personally, I'm a Kerberos fan
The DEFAULT is still DES/crypt. Why not change it
Because Sun guarantees that every piece of software ever written for Solaris will continue to work in new versions. Some would say that they're penalizing the rest of us for those few users still using 10 year old programs. And yet, there's something to be said for "working out of the box".
You're paying your Unix admins anyway. You might as well have them do something more useful than playing NeTrek all day while the system happily chugs along.
Javascript + Nintendo DSi = DSiCade
"When you sit with a nice girl for two hours, it seems like two minutes. When you sit on a hot stove for two minutes, it
No its not that simple. If it were just a database lookup then simply increasing the size of the salt a million fold would be all that was neccessary to foil any pre-computed attack. The important point in this article is that they it only takes 80 minutes to compute the crypts given the salt.
thus one way to defeat this is to hide the salt as follows. The attack requires stealing the hashed password file. Each entry will be of the form
SALT, HASH_code
since the salt is given they could run their computer for 80 minutes and test 209 billion inversions of the hashcode. However if the stored item were instead: SALT2, crypt( SALT1, crypt(passwd)) then you would have to crack the first one to get the salt for the second. now iterate this.
Some drink at the fountain of knowledge. Others just gargle.
If you are interested, you can take a look at a paper I wrote (back in 1997) on the power of cracking with ten 200MHz Pentium Pro machines.
as far as i'm concerned, sun's pam configuration and flexible login setup only applies to solaris 9 and above. after trying to come up with a sitewide md5 solution i found solaris 8 to be terribly difficult to work with. dont get me started on their broken ldap libraries.
several datacenters i work with independently only offer solaris 8 so "why aren't you using the latest sun distro's" falls on my deaf ears. the huge body of vendor supplied software which calls for solaris 8 just makes it worse.
nis implementations that pass these crypt values around the network just makes keeping them inaccessible to users a nightmare.
crypt-for-passwords is one of those "standard unix" methodologies that needs to have already died a horrible death. the original title of this topic was hopefully appropriate.
I am going to go convert two of my physical binary decision devices into a cup of coffee.
Yes, yes, but where did he store them?
/greger
A quarter.
I'm really not all that familiar with the inner workings of cryptography, but it seems like it'd make sense to do more system-specific hashing. (Not in all cases, of course.)
For example, I was looking at the MySQL tables on a site I run, and realized that my password hash there is the same as on other boxes I have accounts on. For example, 5f4dcc3b5aa765d61d8327deb882cf99 is the MD5 sum of "password," anyone with access to a set of passwords could simply skim through looking for this and other well-known hashes.
In many cases, wouldn't it make a fair amount of sense to use machine-specific algorithms for generating password hashes? It could even be something simple, like taking the hash of "$hostname-$username-$password". You could generate all the hashes you wanted, but if you didn't know my username and the hostname, the hash would end up being different. The end result is that, even though I (insecurely, I know) use the same password (and often, the same username), I would have a different password hash on each machine; you'd have to know all three fields before you could brute-force my password.
Is there something I'm overlooking? This idea seems like it'd work out really well; I can't possibly have just described some revolutionary new idea, though.
________________________________________________
suwain_2
But my point still holds. Sun wants to maintain backward compatibility. Period. That's separate from suck-add implementations of the PAM authentication method. And you're lucky. Last time I administrated Solaris boxes (other than my home workstation), I was stuck with Solaris 7. Trust me, that hurt.
nis implementations that pass these crypt values around the network just makes keeping them inaccessible to users a nightmare.
I feel your pain. That's going to be a difficult one to solve though. How does a user authenticate on a network if they can't pass their hash? One could add public key encryption on top of the authentication protocol, but one slip up in key management and the whole deck of cards collapses. After all, how long will it be until we see a worm that uses distributed power to begin generating all possible MD5 hashes? A daunting task to be sure, but certainly not beyond the current state of computing power.
Javascript + Nintendo DSi = DSiCade
In Jaguar (OS X 10.2) and earlier, type:
.
nidump passwd
In Panther (10.3+) it will not show the hashes, and I believe a different algorythm is used anyway.
W
-------------------
This is my SIG. There are many like it, but this one is mine.
Reading the article there is no way that teracrack is going to deal with a strong password, the hash won't be present in it's table.
Regardless of algorithm, the weak passwords will allways be the first to fall. We can all stop using crypt() and start using md5 hashes, but the same techniques can be applied again, and again the first passwords to fall will be the weak ones.
I hate to sound like a Luddite, but technical problems aren't allways best fixed with more technology. The best use of teracrack that I can see, is the same use that it's predecessor had, to identify weak passwords and identify them to the user and admin to ensure that this core problem is addressed.
"Talk minus action equals nothing" - Joey Shithead, D.O.A.
"Talk minus action equals
And yet key loggers and social engineering are still exactly as fast and effective as they were when they were invented...
Think of the children, Moore's children!!!
- Adam L. Beberg - The Cosm Project - http://www.mithral.com/
Not a new concept, but novel given the use of modern computing resources.
:-) wrote software in 1991/1992 to do this: unfortunately sun4 MP's and about a gigabyte of disk were the best we had.
:-). The output was also stored in bitwise numerical order - matching divide/2 matching very fast.
:-).
I (and probably others, I claim no novelty, only an inventive step
Rather than precompute the entire crypt() space, we precomputed the space for well known words (and combinations of those words with different prefix and suffix), because for any individual word, there are only 32 [I think from memory - 5 bits?] combinations that it can crypt to given the random salt that was possible.
Because of available disk space, we couldn't store the entire precomputed output: so we chose to only store the first N bits of it. This was configurable. I cannot remember the exact figure - sure I could dig the code up out of an old CDROM archive
So the password cracker would then mmap() the couple of gigabyte file, then easily find (or not find) a candidate prefix. If it found the prefix, it ran a few trial crypt()'s to ensure an exact match. In practice, because of the lack of diversity in passwords, there were few false candidate matches: so the password cracker had some extremely large hundreds/millions of equivalent cracks per second as a result of mostly just not finding comparisons, and a few trial runs for succesful targets (I think the running rate of the other two popular crackers at that time was about 100K cracks/second).
Anyway - that was a long time ago - fun and games as a student. I still have the source code
In the Bell Labs UTS Ninth Edition (V9 Unix) released in 1986, the crypt command was moved to manual section 6, games, along with trek, bridge, boggle, etc. Crypt (the lib in section 3) still existed, however.
Its stories like this that remind us that Big Brother would chew through any encryption a user might have.
"Oh, 2048 bits? *yawn* We'll have the results for you in a month."
A symmetric key is twice as hard to brute-force for every bit you add. Nobody "brute-forces" asymmetric (public) keys by trying all possible keys; rather, they try to reverse the mathematical process used to construct the public key from the private one. In the case of RSA, this is multiplication, so you have to factor a number. Currently, the fastest public method to do this is the Multiple Polynomial Number Field Sieve; to make something twice as hard with this, you have to add something like 0.3 to the cube root of the number of bits.
RSA keys (and elliptic curves, and DH keys) are much more likely to be severely weakened by improvements in mathematics than say Blowfish or Rijndael keys. If someone improves the method to reduce the constant in the exponent of the NFS (which would probably be extremely difficult), the 1024-bit RSA key could fall easily; otherwise, even Moore's law gives it only a decade or so to live.
I hereby place the above post in the public domain.
I think your statement is redundant, guy.
Occam's razor is the blind faith in the natural selection of least resistance and in universal oversimplification. -- EF
Based on the fact that we're somewhere around breaking RSA-576 and RC5-72. If we assume proportionality:
576 (asymmetric) / 72 (symmetric) = 8
1024 (asymmetric) / x (symmetric) = 8 => x=128
So I'm guessing it's about 2^(128-72) = ~10^16 times safer. Which should be enough for a while too, but not quite that long.
Kjella
Live today, because you never know what tomorrow brings
Looks like its the crypt for crypt() I couldn't resist, someone had to say it.
Simply walk out the building with the entire mainframe!
Karma: It's all a bunch of tree-huggin' hippy crap!
Back in 1994 I wrote the MD5 based crypt() which it seems almost everybody has adopted from FreeBSD by now: *BSD, Linux/GLIBC, Cisco and appearantly even Solaris.
:-)
The important properties of a good crypt() algorithm are still:
1. Input password is not length or charset restricted.
2. The algorithm is complex enough to not lend itself easily to hardware implementations (FPGAs etc).
3. The Salt is big enough that precomputing dictionaries is not feasible.
You will notice that apart from #1, these are not quantified, DES-based crypt() fulfilled #2 and #3 back in its days, but no longer does so. In a similar way, some day we will declare my MD5 based password scrambler as failing one or both of those criteria because password scrambling is not a case of finding "the" algorithm and putting a checkmark in the box, it needs to on periodically evaluated and improved every decade or so.
That is why I put the $1$ prefix on my MD5-based crypt(), so that you can update to a better algorithm when you need to. OpenBSD has already added a couple of stronger SHA based algorithms ($2...) and more can be added in the future.
In the absense of any other "IANA" for this, I would appreciate if you would register your "magic strings" with me so we don't have collisions.
If you're still using DES-based crypt(), switch to MD5 based crypt() now. Don't wait any longer, you are already 9 years late (IMO).
It's even "free as in free beer"
Poul-Henning Kamp -- FreeBSD since before it was called that...