Re: [nmap-svn] r23266 - in nmap: . nselib scripts

[David Fifield <david () bamsoftware com>]

On Mon, May 23, 2011 at 02:43:47PM -0700, commit-mailer () insecure org wrote:
> Author: batrick
> Date: Mon May 23 14:43:46 2011
> New Revision: 23266
>
> Log:
> Set the math.randomseed value in nse_main.lua on behalf of scripts.
> Since Lua uses the C rand and srand functions, which have a static
> seed for the entire program, we don't want scripts doing this themselves.
>
>
> Modified:
>    nmap/nse_main.lua
>    nmap/nselib/mssql.lua
>    nmap/nselib/rpc.lua
>    nmap/nselib/sip.lua
>    nmap/nselib/tns.lua
>    nmap/scripts/dhcp-discover.nse
>    nmap/scripts/dns-fuzz.nse
>    nmap/scripts/nping-brute.nse
>    nmap/scripts/oracle-enum-users.nse
>    nmap/scripts/sip-brute.nse
>    nmap/scripts/smb-brute.nse
>    nmap/scripts/smb-psexec.nse
>
> Modified: nmap/nse_main.lua
> ==============================================================================
> --- nmap/nse_main.lua (original)
> +++ nmap/nse_main.lua Mon May 23 14:43:46 2011
> @@ -96,6 +96,12 @@
>  local math = require "math";
>  local max = math.max;
>  
> +-- Due to heap randomization (on most Operating Systems), we can use a
> +-- Lua function address as a good seed for the C srand function. If there
> +-- is no heap randomization, it's still a decently random integer; that is,
> +-- it's no better or worse than os.time().
> +math.randomseed(tonumber(tostring(function() end):match("function: (0x%x+)")));
> +

This is a good idea, to centralize the seeding. But I'm a bit worried
that an attacker could observe the numbers produced by NSE, invert the
generator, and learn an address in Nmap's memory space to defeat ASLR.
Could we use for example get_random_uint from Nbase instead? That's
ultimately seeded by /dev/urandom.

David Fifield
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/