6 ms·
First published in Ars Technica: http://arstechnica.com/security/2013/05/how-crackers-make-minced-meat-out-of-your-passwords/ http://arstechnica.com/security/2
by jacques_chester 13y ago
First published in Ars Technica:
http://arstechnica.com/security/2013/05/how-crackers-make-minced-meat-out-of-your-passwords/ http://arstechnica.com/security/2013/05/how-crackers-make-mi...
and discussed here:
https://news.ycombinator.com/item?id=5777719 https://news.ycombinator.com/item?id=5777719
The main point to remember was that the story is to expose the tactics and thinking process of attackers, not to be a perfect representation of true life. They deliberately relaxed some constraints (giving attackers unsalted MD5s, for example) but gave a fixed deadline to impose pressure.
- tptacek 13y agoThey used GPU crackers, not rainbow tables. Salting would not have made a difference.
- DuskStar 13y agoActually, salting makes a HUGE difference, even without rainbow tables. Salting does nothing to the difficulty of cracking any individual password, but it makes you crack each person's password INDIVIDUALLY. So you have to crack 123456 + %randomString1%, 123456 + %randomString2%, etc instead of cracking 123456 once for all of the entries on the list. So if it's a list of 1,000,000 password hashes + salt, that will be ~1,000,000 times harder to crack all of them than a simple unsalted list.
- tptacek 13y agoYou don't understand how GPU crackers work, or, for that matter, how Alec Muffet's "Crack" works, from the early 1990s, the original password cracker that everyone used to use to work on Unix password files. Suffice it to say: no, salting does not make a huge difference.
- DuskStar 13y agoOk, I may be confused. I thought that with an unsalted list, you could try a password (generated by your algorithm of choice, whether that be dictionary based or brute force) and compare it to the entire list of hashes, because Alice's "123456" would hash to the same thing as Bob's "123456". But if the list was salted, Alice and Bob would have different stored hashes - and so you would not be able to parallelise cracking across users. Yes, if you only want to crack Alice's password, the salt will not slow you down. But if your goal is to crack a million, wouldn't the salt slow you down? If I'm missing something, I would really like to know. Always good to learn something new!
- cmurphycode 13y agoYour theory of salting-per-user is correct. Rainbow tables are useful if the time it takes to compute is big enough that storing a large list is useful. The GPU crackers are so fast nowadays that that isn't the case. If you can try 500M per second, that's about 10 GB of SHA1 generated! Imagine storing a day's worth of work in a rainbow table :) It's better to use a proper password hash, which will be slow, and will likely have per-user salting built in, anyway.
- mechanical_fish 13y agoRead the Ars Technica article. The hole in your logic is: there is no point in guarding against the case where, if two users share the same password, cracking one will crack the other, because the only time that two users share a password is when that password is really weak. What are the odds that two users share the same password? If passwords were totally random and 24 characters long the answer would be "pretty low". And the odds of cracking one of these passwords by brute force would also be pretty low. If only we lived in that world. In the real world the odds of two users sharing a password are frighteningly high, but that's because most passwords are awful: they are things like "pizzapizza" or even "password". The cracking programs can crack one of these in a millisecond, because they are equipped with lists of thousands of real-world weak passwords harvested from real websites. And then, if another user shares this awful weak password, that user will also get cracked, a few milliseconds later. There is no point in "parallelising cracking across users" because it's so easy to just guess all the weak passwords for every user. Another way to think about it: A salt could only help in cases where (a) two users share the same password, but (b) that password doesn't appear in the list of several million other real-world passwords that is built into the password cracker. And that just isn't worth worrying about.
- Buge 13y agoIf you have a database of 100 unsalted hashed passwords and a dictionary of 10,000 words you could hash the 10,000 words and compare each resulting hash to each hash in the database. So that would take 10,000 hash operations. But if the passwords were individually salted you would have to individually hash each dictionary item for each hash in the database. So that would be 1,000,000 hash operations.
- ssafejava 13y agoCould you explain why? The post below explains about how two users really shouldn't share the same password, and if they do, it's easily crackable - that makes sense. However, if you had a list of 100,000 passwords and did, say, a brute force attack of all passwords with letters, numbers, and special characters to a length of 8, without salted hashes you would be able to run that brute force once and grab every password matching the criteria. With salts, you would have to run it once per salt. Am I off on that?
- jacques_chester 13y agoYou're not wrong, but: 1. If you're using salt, it implies you chose to roll your own key function using hashes. That's a bad idea, because: 2. GPUs can produce so many combinations per second that the difference between salted and unsalted is basically indistinguishable for smart attackers.