Slashdot Mirror
8-Character Windows NTLM Passwords Can Be Cracked In Under 2.5 Hours (theregister.co.uk)
HashCat, an open-source password recovery tool, can now crack an eight-character Windows NTLM password hash in less than 2.5 hours. "Current password cracking benchmarks show that the minimum eight character password, no matter how complex, can be cracked in less than 2.5 hours" using a hardware rig that utilizes eight Nvidia GTX 2080Ti GPUs, explained a hacker who goes by the pseudonym Tinker on Twitter in a DM conversation with The Register. "The eight character password is dead." From the report: It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers. It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. Tinker estimates that buying the GPU power described would require about $10,000; others have claimed the necessary computer power to crack an eight-character NTLM password hash can be rented in Amazon's cloud for just $25.
NIST's latest guidelines say passwords should be at least eight characters long. Some online service providers don't even demand that much. When security researcher Troy Hunt examined the minimum password lengths at various websites last year, he found that while Google, Microsoft and Yahoo set the bar at eight, Facebook, LinkedIn and Twitter only required six. Tinker said the eight character password was used as a benchmark because it's what many organizations recommend as the minimum password length and many corporate IT policies reflect that guidance. So how long is long enough to sleep soundly until the next technical advance changes everything? Tinker recommends a random five-word passphrase, something along the lines of the four-word example popularized by online comic XKCD, "correcthorsebatterystaple." That or whatever maximum length random password via a password management app, with two-factor authentication enabled in either case.
NIST's latest guidelines say passwords should be at least eight characters long. Some online service providers don't even demand that much. When security researcher Troy Hunt examined the minimum password lengths at various websites last year, he found that while Google, Microsoft and Yahoo set the bar at eight, Facebook, LinkedIn and Twitter only required six. Tinker said the eight character password was used as a benchmark because it's what many organizations recommend as the minimum password length and many corporate IT policies reflect that guidance. So how long is long enough to sleep soundly until the next technical advance changes everything? Tinker recommends a random five-word passphrase, something along the lines of the four-word example popularized by online comic XKCD, "correcthorsebatterystaple." That or whatever maximum length random password via a password management app, with two-factor authentication enabled in either case.
Instead of 1 2 3 4 5, it will now be 1 2 3 4 5 6 7 8 9 0
Just BLOCK the acccount after letâ(TM)s say.. 10 wrong tries???
Even if you auto lock the account for 2 hours between each 10 wrong tries, the 2.5 hours of brute force hacking becomes weeks or years of trying...
And it should take a real sysadmin more than a day or two to see his system is being attacked with such alert in log system
So itâ(TM)s a pretty easy fix to do, Microsoft...
It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers.
It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers.
It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers.
Even NTLMv2 is now over 20 years old. It's unsalted, easily parallelizable and you can't adjust the number of hash operations performed. It just can't deal with the modern world. And Microsoft has had tools available for like 5 years now that make it possible to see whether you can disable NTLM, see https://johan.grotherus.com/20... for one writeup. If you have a decently sized environment, this probably won't be easy, but you should start sooner rather than later. As soon as you are able to pull the plug on it, a lot of the easy "pass the hash" attacks become impossible, and those are more dangerous than someone getting to your ntds.dit file in todays age of gratuitous hard disk encryption anyway.
And most people aren't able to create secure passphrases. You need to use completely independant words to actually get a good passphrase, and if someone doesn't understand the information entropy theory behind it, they'll automatically gravitate towards related words. And a passphrase like "housegardengreengrass" has an absolutely abominable complexity of like 20000 * 100 * 100 * 100 or 2^32.
You have to say it three times:
It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers.
Whatever the NIST recommends as the minimum paasword length, my bank sets as the maximum password length.
about every 2.5 hrs.. so again, not much has changed? cease fire stand down,, there's plenty of good reasons to keep us all around.. in the moms we trust.. thanks again
NTLM hashes are disabled by default since Vista.
a few years back, MS Office had a document password protect feature which was cracked instantly by an open source tool.
Slashdot, fix the reply notifications... You won't get away with it...
Couple of questions:
It wasn't entirely clear from the summary /s
Avantslash - View Slashdot cleanly on your mobile phone.
Maybe worth mentioning: https://sites.google.com/site/...
Let us say that Hackerman acquires your password hash and exports it off your computer, across a series of tubes, onto his computer. Consider that now he has an unlimited amount of time in which to solve the hash, and access to almost infinite cheap storage for rainbow tables in order to make a time/storage trade-off. This makes it somewhat ridiculous to consider any palpable difference between 8 characters and 16, at least for these weak hashes
With this being said, the problem is passwords are a bad way to do authentication, not that passwords are too short. You can keep making the passwords more complex and encrypted in a more robust fashion with more salt and pepper, but it doesn't help when I'm just going to write it on a sticky note and leave it on my desk. SQRL or something similar is the inevitable future, see: https://www.grc.com/sqrl/sqrl.htm
Note to author: It was determined during WWII that repeating plaintext makes it far easier for an opponent to crack the cyphertext. Just sayin'.
I have a datacenter full of Hashcat rigs - used to be my crypto mine but I re-purposed and now do fee-based password recovery for corporate and law enforcement clients.
Hashcat is pretty fun and has a scripting language of sorts for narrowing the attack space. If you have knowledge of the corporate password rules you're dealing with (which SIGNIFICANTLY reduce the attack space) it's actually not uncommon to discover even a complex password in a couple of days.
The bottom line is that everyone needs to use stronger passwords, and corporations really need to remove the impediments that reduce attack space.
As an example, let's take a simple example where a keyboard has all the capital and lowercase letters, and numbers 0 through 9. There are 52 possible letters and 10 possible numbers - 62 potential characters. An 8 character password has 62^8 or 218,340,105,584,896 possible combinations.
If I impose a rule that says you must have at least one capital letter, that more than halves the attack space because one combination drops from 62 possibilities to 26, and our new attack space is only 91,561,979,761,408.
If I say you have to have one capital letter and one number, that reduces a combination from 62 to 10, and our new space is only 14,768,061,251,840 passwords.
A GTX 1070 will do a Kerberos 5 password at about 145 million per second, so a single rack of 12 of them will do 1,740,000,000 passwords/second.
That means I can crack 8 characters, one capital letter and one number in a MAXIMUM of 8487 seconds, and that's assuming the correct password is the last one I try. That's less than 2.5 hours.
I have 200 of those racks in my farm, so it takes me longer to set up the job that it takes to completely exhaust that address space: 42 seconds.
So please, corporate America, keep right on with your silly password rules. They only make my job easier and more lucrative.
I can still remember this over many years... but let me Google the link... fond it XKCD
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
No worries. Just require all users to change their password every 2 hours, and problem solved!
>"NIST's latest guidelines say passwords should be at least eight characters long. Some online service providers don't even demand that much."
The example given is an old method and assumes the cracker has access to the stored encrypted password. Then the discussion turns to a wide/broad generalization about ALL password lengths, and web sites were the example. This isn't logical. An 8 character password is way strong enough if you don't have access to the stored data and all you can do is try brute force- which is easily defeated by throwing in delays or limits.
It also depends on the method used to store the passwords, even if you have access to the stored data,
First of all the cracker will need to get the password hashes, if that can be done I think you have more issues than an 8 character password. I cannot comment on how to get windows password hashes (have not be on it in well over a decade), but on a properly locked down and encrypted modern Linux/BSD getting shadow is almost impossible.
At work we are being moved to a minimum 15 character PW which is changed it every 90 days, all I can say is I am real glad I do now work on corporate help desk. 90% of they time the will be getting calls about forgotten passwords. yes people can use sentences but you know as well as I do that is not going to happen for over 80% of the users
it makes it sound like you can just crack any password in 2.5 hours. The technique described requires that someone have the ntlm.dit file. truth is if you have the ntlm.dit or a /etc/passwd file, password dictionaries crack many password in 10 seconds. kinda strange that it takes that long with 6x2080's. hashcat on my 660ti still kicks as$. main thing to crack passwords faster is to only look for numbers and symbols in the last 2 places. so many password end in 01 or 1! no point in looking for letters there.
Tinker recommends a random five-word passphrase, something along the lines of the four-word example popularized by online comic XKCD, "correcthorsebatterystaple." That or whatever maximum length random password via a password management app, with two-factor authentication enabled in either case.
Except that every site has a different maximum number of characters, requires different special characters, some of them don't allow your favorite special characters, etc. So there's no way you can consistently use some complex patterns that you could actually possibly remember.
It's dead at least in the context of hacking attacks on organizations that rely on Windows and Active Directory. NTLM is an old Microsoft authentication protocol that has since been replaced with Kerberos. According to Tinker, it's still used for storing Windows passwords locally or in the NTDS.dit file in Active Directory Domain Controllers.
Jeez, the submissions here are shockingly badly edited, or, I should say, not edited at all. A repeated phrase in the summary, obvious to a 5 year old. BeauHD... hang your head in shame.
Passwords are terrible security. Period. They should never have been widely implemented. All websites, 100% of websites, and all other systems that require passwords should be moved to physical keys.
NIST's latest guidelines say passwords should be at least eight characters long
I tried "at least eight characters long" but it said passwords could not contain spaces.
I now understand why I've been hacked!
"The eight character password is dead."... yes, because the average hacker has a $10,000 rig just to hack individual Windows passwords. That's a silly statement at this point in time. Beyond all that, being an admin and developer at my company I can tell you that the vast majority of people use more than the minimum amount of characters anyway. At any rate, raw password hacking accounts for about 1% of compromises these days... we really need to focus more effort in other areas as well as education of end-users in a manner that they will pay attention to.
I would assume a rainbow table would be faster than bruteforceing
Length is bullshit if you have a low entropy password.
"onceuponatime" will shatter like glass at 13 characters.
"onceuponatime!" will break.
"OnceUponATime!" will break.
"0nc3up0nat1m3" will break.
"emitanopuecno" will break.
If a human is using certain methods to approach their passwords, those methods can be approximated. A single, low entropy mod. Unless you're eager to believe you're the only human who ever thought of writing a password backwards, or swapping leet characters. Then, sure, the tables won't attack towards your never-seen idea. Which is useless to the community at large, as an advisory.
The maximum entropy : easiest recall ratio is abbreviations. ouatiwadasn for the literaturefags, I guess. These should remain resilient until the kits have an overarching knowledge of human dialogue through the ages, are somehow aware of the cultural incident rate of "Charles Dickens", of texts associated with him, of the most prominent strings, leading to attacks on iwtbotiwtwot (and tree of derivatives). I say "somehow aware" but it WILL happen eventually, probably when blackhats get ahold of some nice Google-developed AI loot on all human writings, meant as Skynet-feed.
This will only affect historic abbreviations so grab some novel saying or pop song lyrics and you're set.
so I'm good
If the password isn't protecting anything of value then 1 character will do - for example any site that makes you create an account so you can use it once.
If the attacker is rate limited and is only interested in one account then a 4 digit PIN will do - think bank cards
If the attacker can attack any one of 50,000 employees and is only rate limited per account a pass phrase of 4 words should be used.
If the attacker has the hash of the pass phrase then a pass phrase of 5 words should be used.
If the attacker has the hashes of 50,000+ phrases then a pass phrase of 6 words should be used.
8 random character passwords are useless, they too strong for the rate limited single account, impossible for 50,000 employees to remember and worthless against an attacker with the hash of the password.
You should also fire everyone involved in the 8 character, at least one upper, one lower, one number and one one special character and change your password every 3 months people. After 6 months almost every employee gives up on creating a strong password and uses a common 6 letter English word, capitalizes the first letter, puts in the number 1 and then a '!'. They then increment the number every 3 months.
NTLM, NTLMv2 and yes Kerberos are all HOPELESSLY insecure CHAP based authentication protocols subject to offline brute force campaigns simply by way of an adversary eavesdropping on authentication process. No server hacking required.
Microsoft STILL insists upon using this crap in its current software when secure alternatives are readily available.
The only way authentication works in practice today is by protecting authentication exchanges using PKI... similar in concept to all of the web login forms on present day websites. (phishers paradise)
As for stored credentials... Salting and amplification make password guessing harder.. (persistence of NTOWF is unnecessarily sad) but not in any meaningful way that would do much to limit effective impact in the event of hash table compromise. With hashes for any number of accounts compromised an attacker with access to salted amplified hashes using present day resources is still assured meaningful victory no matter what.
As policy you are way better off treating hashes as plaintext password lists and acting accordingly because effectively that's exactly what they are.
This is wrong. This is a technology driven directive, placing ever greater burdens of responsibility on users in order to make up for technical weaknesses and limitations.
Hey, I've played the game too. "Passwords should be stronger", with all the complexity requirements we add. Ask yourself this though: Where are we headed with complexity and length requirements?
Some day we are going to have access to processors with millions of cores. Some day we are going to have quantum computers. Some day we are going to have DSPs, specialized co-processors in abundance, PLDs, memristors, FPGAs, and more. What does this imply?
We cannot simply keep dumping additional requirements upon the users. If we do, we wind up with passwords that are the equivalent of War And Peace length novels. Already we have issues with password pushback. This has to stop.
Multi-factor authentication is one possible route. Biometrics, ideally as part of an MFA solution, can be an answer. But passwords are already at or near the maximum ask we can make on our users.
So no, I really DGAF what your "killer rig" has, or what your "amazing software" can do. That's not the issue and you are both naïve and short-sighted to think that it is.
Brute-force attacks like these can only work when attackers can access the passwd hashes so their guessing cost comes down to a few machine cycles. This is why /etc/shadow was developed and eventually will become encrypted itself.
When an attacher has to go to the local OS, let alone a remote net, the cost per guess goes up by many (4-10) orders of magnitude. Decent security watchdogs will throttle guessing even further.
ANY 8 char password? Does that include special chars like ©?
If we include those, the number of possibilities increases immensely, to the point that I certainly wouldn't worry if I had such a password.
Most people, even on this thread seem to exclude even the common, easily accessible symbols from their strategy.
"The password you have entered is already in use by user corp-admin@internal.wscorp.com. Please choose another password".
An 8-character password can fit in a 64-bit register.
Proof: 8 x 8-bit = 64-bit
Guys, just upgrade your Active Directory forest functional levels.
As long as the oldest controller is new than 2012, not a problem.
My memory isn't what it used to be but I believe I read something from MS back in the 90's (when I received my MCSE on NT) that the algo used was different when using a 9+ character password. But most people on this thread realize that complexity isn't the secret. It's the length that matters. All of my pwds are maximum length sometimes 25 characters. That will take a very long time to crack. Its not impossible, but the value of the data will be diminished by the time you break it...or I'll have changed it by then anyways.
It's not just the information, it's also the systems theory behind the information theory.
The underlying problem is that so many passwords in the wild get cracked back to plain text. Any paradigm you come up with is vulnerable to machine learning, which can ultimately identify and extract almost any pattern.
The pattern you describe is this: pick a word somewhat randomly (but not one too long or too hard to type), then use that as a seed word to free associate. Your entropy estimate is good and to my eye occupies the Fermi-estimation sweet spot (which is rather large, in a good way).
But the entropy you report is the conditional entropy on having already decided on the password paradigm (in some cases, this isn't even a guess, if you can associate previously cracked passwords to the same user, or specific password policies to the institution).
With enough randomly chosen paradigms to pick among, you could add maybe another 10 bits of true entropy. But people being people, paradigms are about as randomly chosen as social media networks.
Plus, because of all the cracked passwords, we have strong statistical models about paradigm evolution (these are cultural artifacts, for the most part)—at least for the masses who associate themselves with 2nd-rate IT (my own strong password paradigm is only used on sites highly likely to number among the scrypt enlightened).
Not that I'm including simple plug-board scrambling stages in my paradigm model (such as reflecting keystrokes between the left and right hands, moving home position one keyboard position to the right, or one row up).
A lot depends on the yield model of the attacker. If the goal is to crack as many passwords as possible, then you start with the worst of the worst and increment upwards. You would probably never even rise to plug-board stages.
If you have a value model over the accounts, when some accounts are judged to be a thousand or a million times more valuable once cracked, then the high value targets had better not be depending on their password paradigm adding any true security (unless so invented out of whole cloth, you're reluctant to update your password in less than a decade).
I tend to use apg to seed my passwords. I click generate five or ten times until something grabs my eye, and then I tweak it slightly to make it easier to store in short-term memory as I transcribe it from my password keeper to the passbox by hand. My "throw away" passwords are 11 characters, and more important passwords are usually more like 16. I generally estimate my passwords to have at least 50 true bits of entropy, assuming a very efficient search through password paradigm probability space (possibly by an adversary who has already cracked one of my other passwords into plain text). If the password contains spans of alphabetic dictionary fragments, I rarely estimate the fragment as supplying more than about 13 bits regardless of the letter count (or how close it remains to the original apg output).
Fully directed attacks based on a comprehensive model of human entropy management is an awesome superpower.
I'm sure the NSA has been doing this for decades already: nearly every plaintext password they've ever recovered (more than few) has been melded into some giant statistical model. They've likely identified millions of cognitive paradigms by now (from the ones with billions of breaks, down to some with only ten or so exemplars). There's no doubt in my mind they have explored the use of machine learning to squeeze even more out of this heap (though I suspect it was pretty squeezed out, even before machine learning). I also suspect that as this model is refined, they deliberated re-target recalcitrant breaks from ages past, so as to feed these breaks back in
Damn. So THAT'S how he got me.
are in ancient Aramaic -- break that Hashcat!
work out your own salvation in fear and trembling
There is strength in diversity.
Talmudic jibberish. Password entropy is purely about math. The more mathematical possibilities, the better.