rfunk: (Default)
Add MemoryShare This Entry
When I started using computer networks in the early 90s, I had a six-character password. A few years later the recommendation was a minimum of eight characters, and be sure to mix it up with upper/lower-case, numbers, and maybe some punctuation.

Unfortunately way too many people still haven't even gotten that message; as a network administrator (who tries to avoid knowing people's passwords but somehow does anyway since they don't seem to care) I see some amazingly simple passwords.

Even more unfortunately, some software enforces less-than-good passwords by restricting the punctuation or the password length.

But with the guidelines of a decade ago, it's been generally assumed that brute-forcing a password would take many months at least. However, that assumed that the job would be done on a single CPU. Or possibly much faster on a massively-distributed network on CPUs, but that's still a lot of effort for a few passwords, and apparently less interesting or lucrative than searching for aliens or sending spam.

Who expected cracking a password to be done on a graphics card? In less than a week?

So basically the eight-character password has been cracked. But the 12-character password will still present a challenge for a while, even if you don't include the punctuation and numbers and all that.... though you still might be vulnerable to a dictionary attack if you're not careful.
Mood:: 'geeky' geeky
There are 15 comments on this entry. (Reply.)
 
posted by [identity profile] the-shampoo.livejournal.com at 02:45am on 25/10/2007
Generally I glaze over computer related posts, but this one was very engaging. Interesting read. I am somewhat guilty of not making my passwords random enough, I probably ought to.
 
posted by [identity profile] rfunk.livejournal.com at 07:32pm on 26/10/2007
Don't forget long enough! ;-)

I'm glad I managed a computer post that was engaging. I know half the time I write inscrutable stuff that nobody cares about.
 
posted by [identity profile] chronarchy.livejournal.com at 12:47pm on 25/10/2007
Nice of them to do the math for us. I'm reasonably secure, myself, that my passwords are secure enough for me. And the sensitive data that I work with is better secured, really, than my own stuff.
 

Heh

posted by [identity profile] tosk.livejournal.com at 03:55pm on 25/10/2007
At the office we have a modified version of John The Ripper and can uncover about 75% of the network passwords within 24 hours... and we enforce 8 characters Upper Lower and Special Characters.

Passwords are pretty much FAIL at this point. Security will need to move to tokens or biometrics for the serious stuff.
 
posted by [identity profile] rfunk.livejournal.com at 07:43pm on 26/10/2007
Yeah, dictionary attacks can speed things up dramatically, as can having a list of encrypted passwords to test against. (All the "rippers" I've seen ignore the username except as a hint of text that may be in the password, and most real-world crackers don't have the list of encrypted passwords handy.) Combining a dictionary attack with the stream processing would probably give results in seconds or maybe minutes.

However, using longer passwords (actually passphrases) is still viable. Enforcing 12 characters or more might be a starting point, and just saying "passphrase" rather than "password" might encourage people to think in longer terms. Even dictionary attacks are limited by long passwords.

Anyway, really serious stuff should already be using two-factor (or more) authentication, combining passwords with tokens and/or biometrics. Single-factor authentication hasn't been acceptible for serious stuff in a long time.
 
posted by [identity profile] tosk.livejournal.com at 08:09pm on 26/10/2007
I agree with most of what you said, cept the bit about what attackers might have avaliable... Have you poked at the Rainbow Tables?
 
posted by [identity profile] rfunk.livejournal.com at 08:38pm on 26/10/2007
I'm familiar with rainbow tables via the Coding Horror blog. But a rainbow table only works if they have the password database (in the same scheme the rainbow table was built for), same as the various rippers.

Anyway, salting the encrypted password is the key. As one security person put it bluntly regarding rainbow tables: No modern password scheme is vulnerable to them. Unix has had a "modern" scheme since 1976. Of course, Windows doesn't happen to have a modern password scheme.

And once again, long enough passphrases are key; the rainbow table has a finite length, and if it doesn't cover passphrases as long as yours, it can't crack your passphrase.
 
posted by [identity profile] tosk.livejournal.com at 09:03pm on 26/10/2007
*ding* Windows doesn't happen to have a modern password scheme.

But, most corporations run Windows (I use *nix, but still a lot of the major companies run on Windows). On top of that due to the proliferation of HyperActive Directory al lot of UNIX accounts now have uid/pwd in AD. Thus if you hack AD, you get all the UNIX passwords too. Ptacek is right, when talking about coding for websites, where your password scheme can be homemade and designed on sense.

As for length, I agree... my PGP passphrase has been in the double digits since I started using it. But even that is just a matter of time and technology. Salted passwords help, no doubt, securing anything important 2-factor is a must at this point, as you said. In fact, the PCI requirements for corporations now include 2-factor authentication.
 
posted by [identity profile] rfunk.livejournal.com at 10:05pm on 26/10/2007
So basically, it's Windows that's FAIL at this point, not passwords. :-)
 
posted by [identity profile] zenten.livejournal.com at 09:37pm on 08/03/2008
Wouldn't restricting login attempts to a small a number as once per second take care of that problem?

I mean, true it wouldn't work for login attempts that keep on changing usernames (unless you restricted by IP, but that is not practical for many uses), but I don't see those as being a problem as long as your number of users is sufficiently large.
 
posted by [identity profile] rfunk.livejournal.com at 11:54pm on 08/03/2008
Depends on whether the attacker can bypass that restriction. Often attackers can get the (encrypted) password database, and try as often as they want.
 
posted by [identity profile] zenten.livejournal.com at 12:36am on 09/03/2008
Oh. That makes sense then.

How do they typically get it?
 
posted by [identity profile] rfunk.livejournal.com at 12:39am on 09/03/2008
Network-facing services often have holes (bugs) that allow someone to grab arbitrary files from the system. They just have to know how to exploit the problem and grab the file they want.
 
posted by [identity profile] zenten.livejournal.com at 12:42am on 09/03/2008
Ok. I thought for some reason that most systems (at least run by competent sysadmins) didn't have these holes, or at least holes that anyone had found.
 
posted by [identity profile] rfunk.livejournal.com at 12:50am on 09/03/2008
Well first, you have a lot of questionable conditions there, such as "competent sysadmins". :-)
Second, you have to define "anyone". There are certainly holes that are unknown to the good guys, but known to the bad guys.

Also, going back to your earlier question, it's not always possible to limit the number of logins, depending on the protocol being used.

April

SunMonTueWedThuFriSat
        1
 
2
 
3
 
4
 
5
 
6
 
7
 
8
 
9
 
10
 
11
 
12
 
13 14
 
15
 
16
 
17
 
18
 
19
 
20
 
21
 
22
 
23
 
24
 
25
 
26
 
27
 
28
 
29
 
30