blowfish + libxcrypt

calef13 · 2008-01-09 19:58:42

Hi,

I am looking to enabling blowfish passwords using the article in the wiki: http://wiki.archlinux.org/index.php/Blowfish_passwords
But I am a little concerned about installing libxcrypt, I am not very knowledgeable about glibc but I know that you don't want to screw up. The fact that you have to rebuild pam seems a bit extreme, so is there any chance I could screw up glibc in the future after installing libxcrypt. Also do I really need to build the ABS source tree? Can't I just get a PKGBUILD?

EDIT: Just read the wiki ABS page and I understand it all a bit better now. But I would still like to know if I could mess up glibc and would I have to periodically resync the ABS tree and then rebuild glibc every time an updated version is released?

Calef13

Last edited by calef13 (2008-01-09 20:02:30)

cactus · 2008-01-09 20:18:26

why use blowfish passwords?

salted and hashed passwords would seem better to me.
that wiki page dubiously claims that blowfish is more secure than md5. However, blowfish is an encryption algorithm, while md5 is a hashing algorithm. They are entirely different problem domains.

With blowfish, someone *may* be able to reverse the encryption and get your original password. With md5, they *may* be able to find a collision...a different password that would grant them access, but it wouldn't be your password.

md5 collisions still take *quite a while* to find collisions. If you use salted md5, then you defeat most 'rainbow table' style attacks too.

granted, I am not a professional cryptographer... maybe ask on one of the security mailing lists and see what the professionals say.
I believe openbsd uses blowfish, but I am not sure why (i have no context for the reason behind their choice to use that instead of md5, sha1, or whirlpool-1, or even a better block cipher like twofish or serpent).

phrakture · 2008-01-09 20:31:34

One of the big problems with using symmetric encryption for this is that anyone with access to the DB and the proper salt (or whatever else the algorithm needs) is able to figure out what password you used.

Personally, that scares me. If 'John the Malicious Programmer' could find out that I used "hahawut" as my password, he might be able to get into other places I've used the same password.

calef13 · 2008-01-10 19:09:48

phrakture wrote:

Personally, that scares me. If 'John the Malicious Programmer' could find out that I used "hahawut" as my password, he might be able to get into other places I've used the same password.

But surely if you find a collision with your md5 password an attacker can also get access to places where you used the same password too. I guess recovering the actual password you used might reveal other information, like a common link between your other passwords which is a good point. However if you have ever ran a test with john the ripper the cracking blowfish is much slower than cracking md5. But I think your arguments are pretty convincing and I think I'll stick with md5 anyway.

Thanks for the informative replies.

phrakture · 2008-01-10 19:13:27

calef13 wrote:

phrakture wrote:
Personally, that scares me. If 'John the Malicious Programmer' could find out that I used "hahawut" as my password, he might be able to get into other places I've used the same password.
But surely if you find a collision with your md5 password an attacker can also get access to places where you used the same password too. I guess recovering the actual password you used might reveal other information, like a common link between your other passwords which is a good point. However if you have ever ran a test with john the ripper the cracking blowfish is much slower than cracking md5. But I think your arguments are pretty convincing and I think I'll stick with md5 anyway.

You're describing two different scenarios. Yes breaking the algorithm would yield a password in the md5 or sha1 cases, but this is true of ANY algorithm. I am not debating the relative encryption strengths. I am simply stating that *decrypting* is part of the functionality of the symmetric algorithms, so programs *always* know how to do it. It is not a matter of breaking the encryption, it is just the way they work. One-way hashes, on the other hand, do not function in that manner. Knowing the result of the hash is (theoretically) irreversible.

calef13 · 2008-01-10 20:46:59

Firstly, I am not very clear when I write so apologies for any confusion, I didn't mean to drag you into a debate on their merits I don't understand how I am describing two different scenarios, also are you implying these are conflicting scenarios?

Arch Linux

#1 2008-01-09 19:58:42

blowfish + libxcrypt

#2 2008-01-09 20:18:26

Re: blowfish + libxcrypt

#3 2008-01-09 20:31:34

Re: blowfish + libxcrypt

#4 2008-01-10 19:09:48

Re: blowfish + libxcrypt

#5 2008-01-10 19:13:27

Re: blowfish + libxcrypt

#6 2008-01-10 20:46:59

Re: blowfish + libxcrypt

Board footer