iptables recent / more than one exception

[jdow]

On 2012/05/04 01:15, Reindl Harald wrote:
>
>
> Am 04.05.2012 03:10, schrieb jdow:
>> On 2012/05/03 10:57, Reindl Harald wrote:
>>>
>>> Am 03.05.2012 19:46, schrieb Paul W. Frields:
>>>> On Thu, May 03, 2012 at 04:21:20PM +0200, Reindl Harald wrote:
>>>>> iptables -I INPUT -p tcp -i eth0 ! -s $LOCAL_NETWORK -m state --state NEW -m recent --set
>>>>> iptables -I INPUT -p tcp -i eth0 ! -s $LOCAL_NETWORK -m state --state NEW -m recent --update --seconds 1
>>>>> --hitcount  75 -j REJECT --reject-with tcp-reset
>>>>
>>>> Even when you use comma-separated addresses (allowed when not using
>>>> the '!' operator), iptables actually creates separate rules in
>>>> response to the command.  I believe that's what you need to do in this
>>>> situation
>>>
>>> in theory yes
>>> but practically the reject of this rule would be triggered
>>>
>>> a secuity auditor from a customer is whining the he no longer
>>> can make security-scans
>>
>> Ah, wait a minute. If he cannot make security scans neither can
>> anybody else. So defacto his job is finished.
>
> this is the naive view :-)
>
> the reality in business-to-business relationships is that
> such scans are done with standard software like Nessus
> and if this stops the job is not done - if you are
> speaking about a big customer with a policy which requires
> sec-audits yu can not say "it is done"
>
> a real attacker does not need nessus
>
> he plays around with params manually to find sql-injections
> and so on without triggering any rate-control and maybe
> even bypassing mod_security or whatever application fierwall
>
>> For any exception you place into the rules to allow him to scan you must
>> think VERY carefully what it's effects will be. You might accidentally
>> open up the internal network to him leading to a false positive detection
>> from his security scan.
>
> i know this, that is the reason why i like to exclude him only
> from the rate-control as also done with mod_security

So he's not REALLY testing your security profile, is he? He should be
nibbling around the edges, too.

But, then, I note your setting with --recent is not nearly as stringent as
mine. Any given address gets one connection per minute to ssh. That VASTLY
slows down dictionary attacks. Yours is a significant slow down; but, not
so much that somebody could not, as you put it, nibble around the edges to
get in. You have slowed down such attacks, though. That is good.

It would be handy if there was an iptables rule that allowed skipping the
next rule in order if the special rule hit. Alas, I am unaware of such a
trick potential.

{^_^}