[Snort-users] Promiscuous interface hacks?

Matt Kettler mkettler at ...4108...
Thu May 1 15:35:06 EDT 2003


The fact that the interface is in promisc mode is more-or-less irrelevant 
to an attack involving buffer overflows, format strings, off-by-ones, and 
other memory-corruption-to-execute code style attacks.

The ONLY requirement for these attacks to work is that the data you send be 
processed in an unsafe manner. If a snort process sniffs packets off the 
ethernet interface and copies them into a local buffer in an unsafe manner, 
that overflow can be exploited for arbitrary code execution in exactly the 
same manner as any other buffer overflow.

Note that a buffer overflow need not be a plain jane "exec bin/sh over the 
already established tcp session"... That's just a very classic  example 
that's commonly used against standard servers because the code size is 
small. After all, people do attack buffer overflows in UDP based services too.

Admittedly the snort bug wasn't a straightforward strcpy overflow, it was 
an off-by-one, but the attack is still valid and possible. The fact that 
snort isn't involved in a tcp session with you, and is instead sniffing 
packets does not make the attack impossible, just different.



At 09:47 AM 5/1/2003 -0500, Paul Schmehl wrote:
>I admit I'm ignorant of this (which is why I'm asking the question), but I 
>think this is probably the best list that I'm aware of to ask this question.
>
>Is anyone aware of any methods (or white papers describing methods) that 
>describe ways that can be used to hack a box through a NIC that is in 
>promiscuous mode?  I'm curious because I'm wondering how serious the 
>recent vulnerabilities in snort really are to a box that's set up in 
>promiscuous mode.





More information about the Snort-users mailing list