httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From r..@ai.mit.edu (Robert S. Thau)
Subject Re: Digest Authentication
Date Tue, 06 Jun 1995 11:33:01 GMT
   From: guillory@ncsa.uiuc.edu (Stan Guillory)
   Date: Tue, 6 Jun 1995 10:12:27 -0500 (CDT)

   Okay, Brandon. I'll jump in and make a fool of myself.

   First, the form of .htpasswd files does not matter for
   digest authentication. It uses different files. 

Yup, it has to (for reasons we all seem to agree on).  We also all
seem to agree that the server does not have to store the password
itself, but it does have to store H(A1), where (as you say)

   A1 = user-name + ':' + realm + ':' + password

So, the digest-auth files can store something like 

   user:realm:H(A1)

which is (we agree) insufficient to recover the password.  However, I
don't think it's true that...

   To crack this, you would need to know the password. 

...because the H(A1) value, which *is* stored in the .md5passwd file
(or whatever you're calling it) is sufficient to allow a (hacked)
client to generate legitimate-looking requests, even if that client
does not have access to the password itself.  So it still looks to me
like compromise of a .md5passwd file does compromise the server.

   One final question. Brandon, you state that access to the password
   file, .htpasswd, will compromise this system. I don't see how. It is
   not enough to know the md5 algorithm. You must know the password
   to recompute H(A1). If passwords are stored one-way crypted in
   .htpasswd, how does access to .htpasswd help you?

Brandon and I were both using ".htpasswd file" as a generic name for
whatever files are used to store server authentication info, whether
or not they are in the current .htpasswd format.  

   Thanks
   Stan
   NCSA

rst

Mime
View raw message