6 ms·

Eight RTX 4090s Can Break Passwords in Under an Hour

31 points by danboarder 4y ago

raydiatian 4y ago
So what’s the new benchmark for safe password length? I’ve been at 24 random characters for awhile, is this too short for HashCat?
- chick3nman 4y ago
  The math says 12+ if randomly generated by something like a password manager and including uppercase, lowercase, numbers, and symbols. 95^12 is too big for pretty much anyone to crack for a very long time, if ever. There is a small caveat to this regarding what type of hashing is being used though. In nearly all cases, even in the case of relatively weak hashes, 12+ is enough, but there are hashes which truncate or break up inputs. Windows LanMan(LM) is probably the most common offender still seen these days that does this and any LM will always be breakable up to 14 characters due to its underlying algorithm leading to a weak separation of 7 + 7 characters. In the case of Windows hashing a default length of 15+ can help to avoid LM being stored but for legacy systems it may be unavoidable.
  - naikrovek 4y ago
    I don't think the NTLM thing of 7+7 has been true since Windows 2000 and the introduction of Active Directory. it's been a long time, I know that.
    - chick3nman 4y ago
      LM is still this way, the algorithm hasn't changed, it's just not often used/stored except in the case of quite old legacy systems. NT hashes, commonly referred to as NTLM hashes, are indeed not length restricted and do not break your password up into 7 + 7 chunks.
- MontyCarloHall 4y ago
  It says 300 Gh/s for NTLM and 200 kh/s for bcrypt. There are 62^12≈10^21 12 character mixed case alphanumeric passwords. At 3*10^11 hashes/second, this would take ~3*10^9 seconds (~100 years) to brute force. For bcrypt, an 8 character password is fine (it would take ~30 years to brute force). Of course, you have no idea what hash function a breached site may have used, so best to err on the side of caution. That said, avoiding password reuse is far more important than the strength of an individual password (not that it’s unimportant). The vast majority of breaches occur due to credential stuffing attacks on other logins, not from the initial breach itself.
- efficax 4y ago
  if you're using all valid characters (numbers, punctuation, upper and lowercase, etc) then every extra character grows the set of possible passwords exponentially, not linearly. 24 characters is a huge problem space, i don't know exactly how many possible characters there are but let's say it's 50, so that's 50^24 which is 41 digit number, a number so big I don't know what to call it and still in the realm of a password that would take so long to crack you would die before it was finished cracking.
  - hn_go_brrrrr 4y ago
    50^24 is roughly 59.6 duodecillion, if you were curious.
jqpabc123 4y ago
How do you fix this? Quite easy really --- simply limit the number of unsuccessful attempts before locking the account --- either temporarily or indefinitely. Most security focused web sites do this now.
- hotpotamus 4y ago
  You're thinking of brute force attempts over the network which are very slow and as you mention should be subject to tarpits and lockouts. This is if you can get your hands on the hashed passwords; I would think a GPU cluster can do something from millions to trillions of attempts per second depending on the algorithm.
  - jqpabc123 4y ago
    Here is my hashed password: Gx5jp2z8,h Let me know when it has been succesfully decrypted. Or better yet, explain how you can possibly know if a decryption is succesful except by testing it. You can make 10 test attempts before you're locked out for 1 hour. Another 10 attempts before the account is locked indefinitely. The number of attempts per second your GPU cluster can make is no longer relevant.
    - pwg 4y ago
      You are still not understanding how these offline attacks work. Assuming that was a hashed password (it is not), the way that these boxes crack it is they hash all possible passwords and test the resulting hash to see if it matches your hash. When they find an input that matches your hash, they have "decrypted your password". What an offline attack does not do, which is where you are mixed up, is try to log into Facebook/Twitter/etc. over and over trying different possible passwords. Some other security breach reveals the stored hashes of user passwords, and that stolen list of password hashes is what gets run through these systems. Then, when they find a batch of inputs that hash to some of the stolen hashes, then they start trying those both at the site of the breach, and at other sites, often to great success, because they never trip the "10 test attempts" (they have an input that works, first time) and due to the all too common reuse, learning a users craigslist (for example) password is often enough to gain access to their gmail account.
lakomen 4y ago
There is no mention of how long those passwords are so for me this is clickbait
- shellfishgene 4y ago
  Article: "The researchers estimate that a purpose-built password hashing rig (pairing eight RTX 4090 GPUs) could crack an eight-character password in 48 minutes."
  - virodoran 4y ago
    For further clarity, this is for NTLM hashes. And the entire article is indeed clickbait and considering the numbers they're referencing, they likely came from this tweet [1] which the article doesn't link as a source at all. [1] https://twitter.com/tinkersec/status/1580722789245280257 https://twitter.com/tinkersec/status/1580722789245280257

31 points by danboarder 4y ago

So what’s the new benchmark for safe password length? I’ve been at 24 random characters for awhile, is this too short for HashCat?

chick3nman 4y ago
The math says 12+ if randomly generated by something like a password manager and including uppercase, lowercase, numbers, and symbols. 95^12 is too big for pretty much anyone to crack for a very long time, if ever. There is a small caveat to this regarding what type of hashing is being used though. In nearly all cases, even in the case of relatively weak hashes, 12+ is enough, but there are hashes which truncate or break up inputs. Windows LanMan(LM) is probably the most common offender still seen these days that does this and any LM will always be breakable up to 14 characters due to its underlying algorithm leading to a weak separation of 7 + 7 characters. In the case of Windows hashing a default length of 15+ can help to avoid LM being stored but for legacy systems it may be unavoidable.
- naikrovek 4y ago
  I don't think the NTLM thing of 7+7 has been true since Windows 2000 and the introduction of Active Directory. it's been a long time, I know that.
  - chick3nman 4y ago
    LM is still this way, the algorithm hasn't changed, it's just not often used/stored except in the case of quite old legacy systems. NT hashes, commonly referred to as NTLM hashes, are indeed not length restricted and do not break your password up into 7 + 7 chunks.
MontyCarloHall 4y ago
It says 300 Gh/s for NTLM and 200 kh/s for bcrypt. There are 62^12≈10^21 12 character mixed case alphanumeric passwords. At 3*10^11 hashes/second, this would take ~3*10^9 seconds (~100 years) to brute force. For bcrypt, an 8 character password is fine (it would take ~30 years to brute force). Of course, you have no idea what hash function a breached site may have used, so best to err on the side of caution. That said, avoiding password reuse is far more important than the strength of an individual password (not that it’s unimportant). The vast majority of breaches occur due to credential stuffing attacks on other logins, not from the initial breach itself.
efficax 4y ago
if you're using all valid characters (numbers, punctuation, upper and lowercase, etc) then every extra character grows the set of possible passwords exponentially, not linearly. 24 characters is a huge problem space, i don't know exactly how many possible characters there are but let's say it's 50, so that's 50^24 which is 41 digit number, a number so big I don't know what to call it and still in the realm of a password that would take so long to crack you would die before it was finished cracking.
- hn_go_brrrrr 4y ago
  50^24 is roughly 59.6 duodecillion, if you were curious.

jqpabc123 4y ago

How do you fix this? Quite easy really --- simply limit the number of unsuccessful attempts before locking the account --- either temporarily or indefinitely. Most security focused web sites do this now.

hotpotamus 4y ago
You're thinking of brute force attempts over the network which are very slow and as you mention should be subject to tarpits and lockouts. This is if you can get your hands on the hashed passwords; I would think a GPU cluster can do something from millions to trillions of attempts per second depending on the algorithm.
- jqpabc123 4y ago
  Here is my hashed password: Gx5jp2z8,h Let me know when it has been succesfully decrypted. Or better yet, explain how you can possibly know if a decryption is succesful except by testing it. You can make 10 test attempts before you're locked out for 1 hour. Another 10 attempts before the account is locked indefinitely. The number of attempts per second your GPU cluster can make is no longer relevant.
  - pwg 4y ago
    You are still not understanding how these offline attacks work. Assuming that was a hashed password (it is not), the way that these boxes crack it is they hash all possible passwords and test the resulting hash to see if it matches your hash. When they find an input that matches your hash, they have "decrypted your password". What an offline attack does not do, which is where you are mixed up, is try to log into Facebook/Twitter/etc. over and over trying different possible passwords. Some other security breach reveals the stored hashes of user passwords, and that stolen list of password hashes is what gets run through these systems. Then, when they find a batch of inputs that hash to some of the stolen hashes, then they start trying those both at the site of the breach, and at other sites, often to great success, because they never trip the "10 test attempts" (they have an input that works, first time) and due to the all too common reuse, learning a users craigslist (for example) password is often enough to gain access to their gmail account.

lakomen 4y ago

There is no mention of how long those passwords are so for me this is clickbait

shellfishgene 4y ago
Article: "The researchers estimate that a purpose-built password hashing rig (pairing eight RTX 4090 GPUs) could crack an eight-character password in 48 minutes."
- virodoran 4y ago
  For further clarity, this is for NTLM hashes. And the entire article is indeed clickbait and considering the numbers they're referencing, they likely came from this tweet [1] which the article doesn't link as a source at all. [1] https://twitter.com/tinkersec/status/1580722789245280257 https://twitter.com/tinkersec/status/1580722789245280257