Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 11 Dec 2014 11:39:43 +0100
From: Hector Marco <>
Subject: Re: Re: Offset2lib: bypassing full ASLR on 64bit Linux

Hi Mike,

El 09/12/14 a las 02:04, Mike Hommey escribió:
>> Hi,
>> This is a disclosure of a weakness of the ASLR Linux implementation.
>> The problem appears when the executable is PIE compiled and it has an
>> address leak belonging to the executable. We named this weakness:
>> offset2lib.
>> In this scenario, an attacker is able to de-randomize all mmapped
>> areas (libraries, mapped files, etc.) by knowing only an address
>> belonging to the application and the offset2lib value.
>> We have built a PoC which bypasses on a 64 bit Linux system, the three
>> most widely adopted and effective protection techniques: No-eXecutable
>> bit (NX), address space layout randomization (ASLR) and stack smashing
>> protector (SSP). The exploit obtains a remote shell in less than one
>> second.
>> We have proposed the ASLRv3 which is a small Linux patch which removes
>> the offset2lib weakness.
>> Details of the weakness, steps to exploit the offset2lib weakness, a
>> working proof of concept exploit, recommendations and a demonstrative
>> video has been publish at:
> If you have the base address of the executable and are able to read any
> byte in the process address space, as my reading of the paper suggests,
> then you don't even need offsetlib, and no amount of ASLR can save you.

Maybe you misunderstand our paper, we do not rely on been able to read
any byte of the process. Actually, our PoC works by writing bytes on
the stack to de-randomize the PIE executable. This is not new at all,
Packet Storm Security published a paper about how to do it:,-an-example-from-2013.pdf 
(August 2013), One year later, almost the same
technique was published (as new?) in a paper called "Hacking Blind": (May 1014).

The conclusion of the Packet Storm paper was that even de-randomizing 
the process you can do nothing. This is because you can only redirect 
the control flow and that time they didn't know where to jump.

The Offset2lib is a step forward which enables to use (jump to)
the libraries without reading any other data from the executable
as "Hacking Blind" requires. By using the offset2lib to bypass the ASLR, 
the Hacking Blind will be deprecated until offset2lib is fixed.

> Just read the ELF program headers that you can find at the base address
> of the executable to find the PT_DYNAMIC segment. In that segment, find
> the DT_DEBUG entry, which will give you a pointer to a r_debug struct,
> which definition you can find in /usr/include/link.h.
> That struct has a r_map member that gives you a linked list of libraries
> the dynamic linker ( loaded, with the base address for each of
> them.
> Those data structures are used by gdb, so removing the DT_DEBUG pointer
> would break debugging with gdb.
> Mike

Hector Marco.

Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.