Security Basics mailing list archives

Re: Slow down blind SQL injection


From: Simon <simon.xhz () gmail com>
Date: Tue, 9 Oct 2007 23:51:56 -0400

Hi,
  I think it is bad to use the "timing" information of requests to
start counter-attacks (such as slowing down, or sending a message to
the user, etc).  But it could be useful to log such thing or popup an
alert to a human operator.  The alert would say something like
"User1234 session start 3 min ago; delay between requests <1second
(0.343sec); alarm #4 for this user"

If it happens once, you can assume the user is a real human, if it
happens lots you can assume there might be automation tools involved.

Moving fast through HTML forms is not to be considered harmful in any
case, it is most certainly convenient for the user!

HTH,
  Simon

On 10/9/07, Francois Larouche <francois.larouche-ml () sqlpowerinjector com> wrote:
Hi,

I completely agree with Shulman, a user especially if it's an important
one (director and above, or worst a important customer) won't look at
this "special protection" feature that impedes on the normal process of
the application with a good eye. Time is money. And in the other hand a
automated tool won't care about that delay anyways. I know that my tool
doesn't care about time delay, I can just start it and go work on
something else and just be patient. If it takes 2 hours to get the admin
credentials instead of 15 minutes? Who cares, I still got it, no? :) And
even better, now the network administrator won't be alarmed by a cluster
of crazy number of requests made about the same time.

In any cases, by personal experience most of the time if there is a spot
with blind sql injection then the chances are high that somewhere else
there is a place where you can reflect data in much fastest way. (with
UNION or in an sql genered error reflected by the webpage such as or 1
in (SELECT user)) So it defeats all the efforts you put in, and only
succeeded to eventually reduce the user experience.

It's good that you try to find solutions but just beware to not make the
security solution more important than the business. My personal advice
is try to find a solution that will be as transparent as possible to the
user.

Cheers,

Francois
Hi,

I believe this solution is a bit problematic.
Cosider a scenario of a user not remembering the right username or
password, and retyping several times or a user that is not familiar
with a keyboard and inserting typos unintentionally.
Your suggestion is to mistakenly interpret such user as an attacker
performing SQL Injection queries?

In addition an attacker that is determined to hack your site will
tolerate the "slow down" however the user will not tolerate those.

I do not see how much you can profit out of this solution and if you
happen to think of a different alternative please update (sounds like
a good research idea).

Best Regards,
S.H.



From: Tiago Batista <tiagosbatista () gmail com>
To: security-basics () securityfocus com
Subject: Slow down blind SQL injection
Date: Wed, 3 Oct 2007 04:11:30 +0100

Hello all

Today I was barainstorming and came up with an idea that my help slow
down blind sql injection on a web application.

I remembered that usually a user will read a page before subbmiting a
new query, and that takes time, so why not keep a timestamp on the user
session and enforce some time between queries?

I did not search to find out if some applications out there are using
this, but I would like your input on the folowig:

1. depending on the timestamp, do you think the users will be very
anoyed at some error asking them to try again in a few seconds?

2. given that most automated SQL injectors deped on a boolean result
form the query, and this ends up serving a thrid page, how much will
this confuse those tools?

3. Assuming that the pogrammer will log several attempts, will this
help to find and correct blind injection points?

Thank you all

Tiago

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar - get it now!
http://toolbar.msn.click-url.com/go/onm00200415ave/direct/01/






Current thread: