[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: Bright idea...
- From: "Theodore Y. Ts'o" <tytso MIT EDU>
- To: pam-list redhat com
- Cc: pam-list redhat com, pam-list redhat com
- Subject: Re: Bright idea...
- Date: Mon, 24 Jun 1996 14:10:21 -0400
Date: Mon, 24 Jun 1996 12:41:44 -0500 (CDT)
From: Aleph One <aleph1@dfw.net>
The problem with this approach is that the pam_delay module does not know
how long the other modules took and therefore cant calculate how long to
delay so that all of the chain of modules always takes the safe amount
of time and gives away information that might be usefull of an intruder.
In many, if not most, cases this type of timing attack really isn't
practical. Hence, I'm not convinced that it's really worth the huge
increase of complexity to protect against this sort of thing. (In
general the only information that can be leaked is whether the username
or password is incorrect --- and in many cases there's many other ways
to determine whether the username is valid on a system, and in even more
cases users don't care about trying to keep it a secret whether of not a
username is valid.)
If you really care about this, though, there's a much simpler way of
accomplishing it, which is to make the delay module have an option for
delaying a random amount of time. Just have two control knobs --- the
average amount of delay, and the standard deviation of the delay. If
these variables are set appropriately, it will thwart someone who might
figure out (oh no!!!) whether or not a login was rejected because the
username or the password is incorrect.
- Ted
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
[]