Subject: [Phrack] Scanning for RPC Services
---[ Phrack Magazine Volume 7, Issue 51 September 01, 1997, article 10 of 17
-------------------------[ Scanning for RPC Services
--------[ halflife <halflife@infonexus.com>
Remote Procedure Language is a specification for letting procedures be
executable on remote machines. It is defined in rfc1831. It has a number of
good traits, and if you run SunOS or Solaris, you are almost required to make
use of it to some degree.
Unfortunately, there are vulnerabilities in some RPC services that have
caused many machines to be penetrated. Many administrators block access to
portmapper (port 111) in an effort to deny external users access to their weak
RPC services.
Unfortunately, this is completely inadequate. This article details how
trivial it is to do a scan for specific RPC program numbers. The scan can be
performed relatively quickly, and in many cases will not be logged.
First, a little information about RPC itself; when I refer to RPC, I am only
referring to ONC RPC, and not DCE RPC. RPC is a query/reply-based system. You
send an initial query with the program number you are interested in, the
procedure number, any arguments, authentication, and other needed parameters.
In response, you get whatever the procedure returns, and some indication of
the reason for the failure if it failed.
Since RPC was designed to be portable, all arguments must be translated into
XDR. XDR is a data encoding language that superficially reminds me a little
bit of Pascal (at least, as far as strings are concerned). If you want more
information on XDR, it is defined in rfc1832.
As you probably surmised by now, RPC programs are made up of various
procedures. There is one procedure that always exists, it is procedure 0.
This procedure accepts no arguments, and it does not return any value (think
void rpcping(void)). This is how we will determine if a given port holds a
given program, we will call the ping procedure!
So now we have a basic idea on how to determine if a given port is running
a given RPC program number. Next we need to determine which UDP ports are
listening. This can be done a number of ways, but the way I am using is
to connect() to the port and try write data. If nothing is there, we
will (hopefully) get a PORT_UNREACH error in errno, in which case we know
there is nothing on that port.
In the given code, we do a udp scan, and for every listening udp port, we
try to query the ping procedure of the program number we are scanning for.
If we get a positive response, the program number we are looking for exists
on that port and we exit.
Vova
