Per-Analysis Network Routing
2.0-rc1 it is possible to feature per-analysis network
routing. In other words, if you have one VM and three samples to analyze, it
is possible to deny internet access for the first analysis, route the second
analysis through a VPN, and pull the third analysis through the Tor network.
However, aside from the more advanced per-analysis routing, it is naturally also possible to have one default route - a setup that used to be popular before, when the more luxurious routing was not yet available.
In our examples we'll be focusing on
VirtualBox as it is our default
Simple Global Routing
Before delving into the more complex and feature-rich per-analysis network
routing we'll first cover the older approach, which is based on global
iptables rules that are, once set, not changed anymore.
In the following setup we're assuming that the interface assigned to our
VirtualBox VM is
vboxnet0, the IP address of our VM is
/24 subnet), and that the outgoing interface connected to the
eth0. With such a setup, the following
iptables rules will
allow the VMs access to the Cuckoo host machine (
192.168.56.1 in this
setup) as well as the entire internet as you would expect from any application
connecting to the internet.
$ sudo iptables -t nat -A POSTROUTING -o eth0 -s 192.168.56.0/24 -j MASQUERADE # Default drop. $ sudo iptables -P FORWARD DROP # Existing connections. $ sudo iptables -A FORWARD -m state --state RELATED,ESTABLISHED -j ACCEPT # Accept connections from vboxnet to the whole internet. $ sudo iptables -A FORWARD -s 192.168.56.0/24 -j ACCEPT # Internal traffic. $ sudo iptables -A FORWARD -s 192.168.56.0/24 -d 192.168.56.0/24 -j ACCEPT # Log stuff that reaches this point (could be noisy). $ sudo iptables -A FORWARD -j LOG
And that's pretty much it, with these rules set we're almost good to go. However, these rules won't be doing any packet forwarding unless IP forwarding is explicitly enabled in the kernel. To do so, there is a temporary method that survives until a shutdown or reboot, and a permanent method that is taken into account when booting the machine. Simply put, generally speaking you'll want to run both commands:
$ echo 1 | sudo tee -a /proc/sys/net/ipv4/ip_forward $ sudo sysctl -w net.ipv4.ip_forward=1
Iptables rules are not persistent between reboots, so if want to keep
them you should use a script or just install
Newer Linux distributions have adopted udev's interface-naming scheme. It's
important to note that this means
eth0 may no longer be your primary
interface. Possible interface names include
emXX, where the
XX part identifies a number. This is particularly
important to note for the NAT statement above.
Per-Analysis Network Routing Options
Having discussed the old school method for routing analyses through a network interface we will now walk through the dynamic network routing components that allow for much more granular network routing.
As outlined in the introduction for this chapter of the documentation it has
been possible since Cuckoo
2.0-rc1, when we introduced the :ref:`rooter`,
to do per-analysis network routing. Since then various bugs have been resolved
and more network routing options have been added.
Following is the list of available routing options.
|:ref:`routing_none`||No routing whatsoever, the only option that does not require the Cuckoo Rooter to be run (and therefore also the default routing option).|
|:ref:`routing_drop`||Completely drops all non-Cuckoo traffic, including traffic within the VMs' subnet.|
|:ref:`routing_internet`||Full internet access as provided by the given network interface (similar to the :ref:`simple_global_routing` setup).|
|:ref:`routing_inetsim`||Routes all traffic to an InetSim instance - which provides fake services - running on the host machine.|
|:ref:`routing_tor`||Routes all traffic through Tor.|
|:ref:`routing_vpn`||Routes all traffic through one of perhaps multiple pre-defined VPN endpoints.|
Using Per-Analysis Network Routing
Having knowledge about the available network routing options it is time to actually use it in practice. Assuming Cuckoo has been configured properly taking advantage of its features is really as simple as starting the Cuckoo Rooter and choosing a network routing option for your analysis.
Documentation on starting the
Cuckoo Rooter may be found in the
For Linux kernel TCP/IP source routing reasons it is required to register each
of the network interfaces that we use with
iproute2. This is trivial, but
As an example we'll be configuring :ref:`routing_internet` (aka the
dirty line) for which we'll be using the
eth0 network interface -
reverting back to Ubuntu 14.04 and older terminology here for a second (Ubuntu
16.04 uses network interface names based on the hardware manufacturer, as you
will likely have seen happen on BSD-based systems since forever).
eth0 we're going to open the
/etc/iproute2/rt_tables file which will look roughly as follows:
# # reserved values # 255 local 254 main 253 default 0 unspec # # local #
Now roll a random number that is not yet present in this file with your dice
of choice and use it to craft a new line at the end of the file. As an
iproute2 could look as follows:
# # reserved values # 255 local 254 main 253 default 0 unspec # # local # 400 eth0
And that's really all there is to it. You will have to do this for each network interface you intend to use for network routing.
The default routing mechanism in the sense that Cuckoo allows the analysis to
route as defined by a third party. As in, it literally doesn't do anything.
One may use the
none routing in conjunction with the
drop routing option is somewhat like a default :ref:`routing_none`
setup (as in, in a machine where no global
iptables rules have been
created providing full internet access to VMs or so), except that it is much
more aggressive in actively locking down the internet access provided to the
drop routing the only traffic possible is internal Cuckoo traffic and
DNS requests or outgoing
TCP/IP connections are blocked.
By using the
internet routing one may provide full internet access to VMs
through one of the connected network interfaces. We also refer to this option
dirty line due to its nature of allowing all potentially malicious
samples to connect to the internet through the same uplink.
It is required to register the dirty line network interface with iproute2 as described in the :ref:`routing_iproute2` section.
For those that have not heard of InetSim, it's a project that provides
fake services for malware to talk to. In order to use
InetSim routing one
will have to setup InetSim on the host machine (or in a separate VM) and
configure Cuckoo so that it knows where to find the InetSim server.
The configuration for InetSim is self-explanatory and can be found as part
$CWD/conf/routing.conf configuration file:
[inetsim] enabled = yes server = 192.168.56.1
In order to quickly get started with InetSim it is possible to download
the latest version of the REMnux distribution which features - among many
other tools - the latest version of InetSim. Naturally this VM will
require its own static IP address which should then be configured in the
routing.conf configuration file.
Although we highly discourage the use of Tor for malware analysis
- the maintainers of
Tor exit nodes already have a hard enough time
keeping up their servers - it is in fact a well-supported feature.
First of all Tor will have to be installed. Please find instructions on installing the latest stable version of Tor here.
We'll then have to modify the
Tor configuration file (not talking about
Cuckoo's configuration for Tor yet!) In order to do so, we will have to
provide Tor with the listening address and port for TCP/IP connections and UDP
requests. For a default
VirtualBox setup, where the host machine has IP
192.168.56.1, the following lines will have to be configured in
TransPort 192.168.56.1:9040 DNSPort 192.168.56.1:5353
Don't forget to restart Tor (
/etc/init.d/tor restart). That leaves us with
the Tor configuration for Cuckoo, which may be found in the
$CWD/conf/routing.conf file. The configuration is pretty self-explanatory
so we'll leave filling it out as an exercise to the reader (in fact, toggling
enabled field goes a long way):
[tor] enabled = yes dnsport = 5353 proxyport = 9040
Note that the port numbers in the
$CWD/conf/routing.conf files must match in order for the two to interact
Last but not least, it is possible to route analyses through a number of VPNs. By defining a couple of VPNs, perhaps ending up in different countries, it may be possible to see if potentially malicious samples behave differently depending on the country of origin of its IP address.
The configuration for a VPN is much like the configuration of a VM. For each
VPN you will need one section in the
file detailing the relevant information for the VPN. In the configuration the
VPN will also have to be registered in the list of available VPNs (exactly
the same as you'd do for registering more VMs).
Configuration for a single VPN looks roughly as follows:
[vpn] # Are VPNs enabled? enabled = yes # Comma-separated list of the available VPNs. vpns = vpn0 [vpn0] # Name of this VPN. The name is represented by the filepath to the # configuration file, e.g., cuckoo would represent /etc/openvpn/cuckoo.conf # Note that you can't assign the names "none" and "internet" as those would # conflict with the routing section in cuckoo.conf. name = vpn0 # The description of this VPN which will be displayed in the web interface. # Can be used to for example describe the country where this VPN ends up. description = Spain, Europe # The tun device hardcoded for this VPN. Each VPN *must* be configured to use # a hardcoded/persistent tun device by explicitly adding the line "dev tunX" # to its configuration (e.g., /etc/openvpn/vpn1.conf) where X in tunX is a # unique number between 0 and your lucky number of choice. interface = tun0 # Routing table name/id for this VPN. If table name is used it *must* be # added to /etc/iproute2/rt_tables as "<id> <name>" line (e.g., "201 tun0"). # ID and name must be unique across the system (refer /etc/iproute2/rt_tables # for existing names and IDs). rt_table = tun0
It is required to register each VPN network interface with iproute2 as described in the :ref:`routing_iproute2` section.