ProFTPD: Firewalls, Routers, and NAT

Basic NAT information
NAT (Network Address Translation) is a system that acts like a proxy, but on a "packet" level. When a computer on your local network connects to a computer on the Internet, the NAT replaces the "from" information of packets with its own address, making your local network invisible to the Internet. Many firewalls perform NAT duties as well, so the following information is valid in firewalled environments as well.

For server systems, NAT can improve security and enable multiple servers to be accessed using a single IP address. This is done by allowing certain ports forwarded "inward" to the local network. However, the part of the FTP protocol known as "passive" data transfers is not by default compatible with NAT solutions. But NAT functionality is possible with ProFTPD versions 1.2rc2 and later.

Note: for details on NAT configuration for Linux, read the Linux IP-masq HOWTO at:
or search for information concerning your OS of choice.

Configuring ProFTPD behind NAT
First configure your installed proftpd so that it works correctly from inside the NAT. There are example configuration files included with the source. Then add the MasqueradeAddress directive to your proftpd.conf file to define the public name or IP address of the NAT. For example:

  MasqueradeAddress  # using a DNS name
  MasqueradeAddress      # using an IP address
Now your proftpd will hide its local address and instead use the public address of your NAT.

However, one big problem still exists. The passive FTP connections will use ports from 1024 and up, which means that you must forward all ports 1024-65535 from the NAT to the FTP server! And you have to allow many (possibly) dangerous ports in your firewalling rules! Not a good situation. For a good description of active versus passive FTP data transfers, see:
To resolve this, simply use the PassivePorts directive in your proftpd.conf to control what ports proftpd will use for its passive data transfers:
  PassivePorts 60000 65535	# These ports should be safe...
Note that if the configured range of ports is too small, connecting clients may experience delays or be completely unable to operate when they request passive data transfers. When the daemon cannot use one of the ports in the configured range, it will fall back to using a kernel-assigned port, and log a message reporting the issue. The clients' ability to use this non-configured port will then depend on any NAT, router, or firewall configuration.

Now start the FTP daemon and you should see something like: - Masquerading as '' (
in the log files.

A Linux Example
This example is for Linux kernel version 2.2.x with ipchains and ipmasqadm. The examples below assume that your FTP server has local address

First we need to enable NAT for our FTP server. As root:

  $ echo "1">/proc/sys/net/ipv4/ip_forward
  $ ipchains -P forward DENY
  $ ipchains -I forward -s -j MASQ
Now we load the autofw kernel module and forward ports 20 and 21 to the FTP server:
  $ insmod ip_masq_autofw
  $ ipmasqadm autofw -A -r tcp 20 21 -h
Then we forward ports for passive FTP transfers. In our proftpd.conf file we restricted passive transfers to ports 60000-65535, so that is what we use here as well:
  $ ipmasqadm autofw -A -r tcp 60000 65535 -h

If instead your Linux system uses IP Filters, then you might do something like the following. First, update your ipf.conf with:

  # Allow passive FTP transfers from ports 49152 to 65534, the IANA-registered
  # ephemeral port range.
  pass in quick proto tcp from any to any port 49151 >< 65535 flags S keep state
Then make sure that the changes take effect by using:
  $ ipf -Fa -f /path/to/ipf.conf

Double Checking
Setting up proftpd that allows passive data transfers srequires that a range of ports be forwarded from the NAT to the local network. This could be a security hazard, but since you can specify what port range to use, you are still able to setup relatively tight firewalling rules. To be sure that you have no other processes listening on the ports you have specified for passive transfers, use a port scanner such as nmap:

  $ nmap -sT -I -p 60000-65535 localhost
If the result says something like:
  All 5536 scanned ports on localhost ( are: closed
then you should be safe.

Frequently Asked Questions

Question: How do I know if my MasqueradeAddress and PassivePorts configuration is working?
Answer: When performing a passive data transfer, an FTP client sends the PASV command to the FTP server. The server responds with the address and port to which the client should connect. For example:

  227 Entering Passive Mode (127,0,0,1,19,6).
The address and port are contained in the parentheses, formatted as a1,a2,a3,a4,p1,p2, where the IP address is:
and the port number is:
  p1 * 256 + p2
If the address seen in the server's response is not a public IP address or the port is not in the port range configured by your PassivePorts, double-check your proftpd.conf. Non-public IP addresses are defined by
RFC 1918, and include 10.x, 172.16.x and 192.168.0.x.

Question: I am using the PassivePorts directive, but my FTP client still doesn't work. In the proftpd debug logging, I see:

  May 20 17:00:55 proftpd[10078] (::ffff:[::ffff:]): Refused PORT 192,168,1,2,193,116 (address mismatch)
Answer: The PORT there means that the FTP client is requesting an active data transfer; this means that proftpd is being asked to actively connect to the client (to the given address, port 49524).

The first problem is, as the log message indicates, that the IP address given in the PORT command,, does not match the IP address from which the client connected (i.e. By default, proftpd will refuse to create a data transfer connection to anywhere except back to the requesting client's IP address; see the AllowForeignAddress directive for more information.

The next problem is that IP address is not a publicly routable I address; specifically, it is part of an RFC 1918 address space. This means that is not possible for proftpd to connect to that address (unless proftpd is located in the same LAN).

The solution for these situations is to a) configure proftpd to use both PassivePorts and MasqueradeAddress, and b) to configure your FTP client to request passive data transfers (via the PASV and EPSV commands), rather than active data transfers.

Question: I am using MasqueradeAddress in my config, using the external DNS name for my site, but proftpd still returns the internal/LAN IP address in the PASV response. Is it a bug?
Answer: Most likely not. If you use something like:

and see the internal IP address in the PASV response, it suggests that proftpd, when starting up, resolves that DNS name from within your LAN, and gets that internal IP address. Rather than using the DNS name, you should explicitly use the external IP address in your MasqueradeAddress directive:
where "" is your real external IP address.

Question: Can I configure proftpd so that it refuses to handle passive transfers?
Answer: If you are using a version of proftpd older than 1.2.10rc1, no. In 1.2.10rc1, support for placing limits on the PASV and PORT (and their IPv6 equivalents EPSV and EPRT) was added, so that you could do the following:

  <Limit EPSV PASV>

Question: How can I make proftpd use a different MasqueradeAddress based on the address of the connecting client?
Answer: This question usually arises in the case where FTP clients connecting from inside the LAN see the same MasqueradeAddress as external clients, which causes problems. That MasqueradeAddress may be necessary in order to allow external FTP clients to do passive data transfers. The internal clients do not need it. To handle this, create a <VirtualHost> section in your proftpd.conf to handle the LAN address of the FTP server, the address that the internal clients are contacting. In this <VirtualHost> section, make sure there is no MasqueradeAddress directive. This way, the external FTP clients "see" the configuration with the MasqueradeAddress they need, and the internal FTP clients "see" a different configuration, one with no MasqueradeAddress.

For those that need to see a concrete example configuration of this:

  ServerName "Some Server Name"
  PassivePorts 60000 65535

  # Note that your LAN address should be used here
    ServerName "Some Other Server Name"

    # Note that there is no MasqueradeAddress directive
    # used in this section!

On the other hand, there can be cases where you really do need different MasqueradeAddress values, in the same vhost, depending on the client IP address. For these situations, you should use classes and the mod_ifsession module, like so:

  <IfModule mod_ifsession.c>
    # Define a class for internal clients
    <Class internal>

    # Internal clients get a different MasqueradeAddress
    <IfClass internal>

    # All other clients get some different, public MasqueradeAddress
    <IfClass !internal>

Contributor: Tobias Ekbom <tobias at vallcom dotcom>

© Copyright 2017 The ProFTPD Project
All Rights Reserved