Firewall Configuration and Testing.

pdf (Size:
204.72 KB / Downloads: 78)

Presented By:
Mohamed Amer, Qamar Nazir
Project Report for Information Security Course
Linköpings universitetet, Sweden

h

Firewall is one of the important security
software that protects your system from other
network. Well configured firewall gives
confidence to the administrator that his system
is protected from malicious attacks. Realizing
good firewall configuration is a matter of testing.
In this project paper we will configure and test a
firewall, and discover how costly can
the testing process be. We also present some
technologies, environments and a way of
doing firewall configuration and testing using
some of the basic rules and included these
results in our report.


 
c
Firewall is the set of rules which protects the
networked computer system from unauthorized
access. It can be implemented on software or
hardware or both.
All messages entering or leaving to system is
through firewall. It allows accessing only those
messages which are allowed by the firewall and
block the others which are not allowed to enter.
Firewall can defend your one system from whole
network or the whole network from the other
network.
There is special security criteria are defined in
the firewall. So every message enter to the
system must meet that specified security criteria
to enter that system.
In this paper we will explain how we will
configure firewall on LINUX environment. After
completing the configuration at first level we will
test it on the same system, and then to the other
systems. This report will contain the results of
our tests and some of our experiences.
Further in section 2 represent the background of
the project, section 3 represent some practical
work, section 4 represent some of the related
work in firewall implementation and finally
section 5 conclude the whole report.



Now a day's Internet is a dangerous place for
your computer but it is not so few years before.
Few years before people can happily do their
business on the web without any protection.
There is very less chance of virus attack,
malware or hacking. Now it is impossible.
Increasing number of Internet users also
increase number viruses and malware. Many of
these don't need permission to execute on your
system.
To protect your computer from today Internet
environment: your pc should have effective
antivirus, antivirus updates and firewall. Linux
operating system includes firewall protection to
enhance security. Properly configured firewall
can definitely increase the security of a network.
This report will cover using a Linux computer as
a firewall between private network, a DMZ
network and Internet. Iptables is the firewall
implementation that is used throughout the
configuration and testing process.



Firewall testing is very important, it gives
confidence to the administrator that the firewall
rules they write are working properly. The right
packet is accepted and right packet is dropped.
Firewall testing is difficult because there are
many parameters which resulting is huge
number of possible parameter combination.
There are number of possible combination
of test cases can be used to test firewall rules.
Typically each test case is viewed as a row in a
table or in database term, a relation. So the
main problems are:
1. Test case selection: which test cases should
be applied? Different mathematical calculation
can be used for it.
2. Test case execution: After selecting test
cases, what method should be applied to
execute it?
For first problem an efficient algorithm will be
sufficient. For second there are libraries to factor
out the many details of packet generation,
transmission and reception [1].
Yong Du and Daniel Hoffman have presented
tools and techniques for testing iptables, the
methods and tools also apply to other firewall
products.
Iptables are used for packet filtering based on
header fields e.g., IP address, TCP and UDP
port and TCP flag. Four main features of
iptables are stateless filtering, state full filtering,
network address/port translation, and
logging.
The syntax of iptables is simple. Typically rule is
ACCEPT or DROP. The rules work like C switch
statement. If Pi matches, then ai is invoked. If
no predicate matches, default action is make
(ACCEPT or DROP). Usually default DROP is
chosen for security reason. Iptables are
implemented using Linux command line.
Testing configuration consist of two PCs- the
driver and system under test (SUT)- The
SUT configured so that traffic enter from eth1
with destination IP address and routed to eth2
and vice versa. ARP (Address Resolution
Protocol) packets are used to map IP (Internet
Protocol) address to MAC (Medium Access
Control) addresses.
The first issue is how to generate, send, and
receive frames on the drivers eth1 and eth2
interface. They have developed a raw socket
library which makes the resulting test
cases much easier to understand and modify.
So at the end result is a simple
open/close/read/write interface, much like the
Linux raw I/O interface.
They created test template by providing different
combination of parameters. There are three
strategies they used for Tuple generation:
Cartesian product generation, boundary value
generation and pair wise generation.
The testing framework for iptables has been
implement in a tool called PBit (Pattern Based
iptables tester). The useful feature of PBit is that
you can modify the test configuration at run time
[1].



In this section we will describe
the testing method implementation, like
Programming language or a scripting language
that is used to implement an automated test
method

The test was performed using the nmap tool by

executing manually a set of nmap scans and
collecting results. The types of scans [6]
performed were:
¢ TCP scan
¢ SYN scan
¢ FIN scan
¢ Null scan
¢ Xmas tree scan
¢ UDP scan
¢ IP protocol scan
¢ ACK Scan
¢ OS Fingerprinting

¢ Window Scan
¢ RPC Scan
¢ List Scan
¢ Version Detection

Each scan command was executed for each

network address space, the DMZ address space
and the LAN address space.




In this section we will describe the technologies
that are using in the project environment like
UML "User Mode Linux", MLN tool and iptables
So we will start with the "UML". UML is a port of
Linux to the Linux system call interface, and
allows users to run any number
of virtual systems (UML instances) without the
need for special privileges. The "UML" system
also includes basic facilities for
networking virtual machines. [4]
To simplify setup of networks of UML instances,
a tool called "MLN" is use. MLN (Manage Large
Networks) is a virtual machine administration
tool designed to build and run virtual machine
networks based on Xen, VMWare Server and
User-Mode Linux. It is ideal for
creating virtual network labs for
education, testing, hosting or simply playing
around with virtual machines. The goal is to
ease the configuration and management
of virtual networks. Xen and User-Mode Linux
are widely used as tools for testing, learning
and virtual hosting. MLN builds and configures
file system templates based on its descriptive
and easy programming language and stores
them in an organized manner. It also generates
start and stop scripts for each virtual host,
enabling you to manage a
running virtual network by stopping
individual virtual machines within a network and
starting them again. MLN makes it possible to
have several separate networks, projects, at
once and even connect them together to create
larger networks. [5]
Ipchains are the most common firewall/Nat
packages running on linux. Iptables are the
enhanced product of ipchains by net filter
organization.
Iptables is packet filtering firewall software.
Packets inspected by iptables are passing
through sequence of rules.
There are total three types of tables.
¢ Mangle table
¢ Filter queue
¢ Nat queue
Mangle table is responsible for alteration of
quality of service bit in TCP header. This is very
rarely used in satellite offices and home offices
(SOHO) environment. Filter queue is
responsible for packet filtering. It contains three
built-in chains where we can put rules. These
are:

Forward chain, Input chain and output chain.

Nat queue contains two built-in chains. Which
are pre-routing and is responsible for Network
address translation. It post-routing? [2]

Queue Type Queue Function Packet

Transformation Chain in Queue Chain Function
Mangle TCP header modificatio n
PREROUTING
POSTROUTING
OUTPUT
INPUT
FORWARD Modification of the TCP packet
quality of service bits before routing occurs.
(Rarely used in SOHO environments)
Filter Packet filtering FORWARD Filters packets
to servers accessible by another NIC on the
firewall.
INPUT Filters packets destined to the firewall.
OUTPUT Filters packets originating from the
firewall
Nat Network Address Translation
PREROUTING Address translation occurs
before routing. Facilitates the transformation of
the destination IP address to be compatible with
the firewall's routing table. Used with NAT of the
destination IP address, also known as
destination NAT or DNAT.
POSTROUTING Address translation occurs
after routing. This implies that there was no
need to modify the destination IP address of the
packet as in pre-routing. Used with NAT of the
source IP address using either one-to-one or
many-to-one NAT. This is known as source
NAT, or SNAT.
OUTPUT Networks address translation for
packets generated by the firewall. (Rarely used
in SOHO environments)
Table 1. Processing For Packet Routed By the
Firewall

For every firewall rule, user needs to specify

iptables and ipchain. As most of the rules are
related to filtering, so if user mentions any rule
without an associated table then it will be
considered as a part of the filter table. So we
can say filter table is a default table.
In the figure 2 [2] packets arrives from Network
and handled by the firewall to create a data
connection.
The packet is first examined by the mangle table
PREROUTING chain, then it is pass through nat
table PREROUTING chain to check, it DNAT of
not. Then its sent for routing. If it is defined to
protected network then it passes through.
FORWARD chain of mangle table for quality
purpose then it is filtered by the rules of the filter
table in its forward chain and if necessary it
goes for
SNAT in POSTROUTING chain of mangle table
and
then it arrives to network B. If the destination
wants to apply, it will follow the same sequence.
If packet is passed through the firewall then it
through INPUT chain of the mangle table, then it
is filtered by a INPUT chain of the filter table.
Then it passes to the firewall application for
some processing.
If firewall reply then packet is sent for the routing
and it is inspected by OUTPUT chain of the
mangle table, if any.




h 


h
is required then OUTPUT chain rule of filter
table are applied to that packet.
Finally POSTROUTING chain of mangle checks
the QoS of packet and then POSTROUTING
chain of nat
table check SNAT of the packet.

 !
"
#


The environment on which we have performed
the firewall configuration and testing was a linux
workstation that has virtual network of virtual
computers that are managed by the MLN tool.
The network schema is described in the lab
document. The tests are done using external
and internal UML instances.
Aggressive OS guesses: T-Home Speedport W
501V WAP (96%), Linux 2.6.12 - 2.6.20 (94%),
Linux 2.6.9 -2.6.26 (94%), Linux 2.6.18 - 2.6.22
(93%), DD-WRT
v24 (Linux 2.6.22) (93%), Linux 2.6.22 - 2.6.23
(93%),
Linksys WRT300N wireless broadband router
(92%), Linux 2.6.19 - 2.6.24 (Gentoo) (92%), D-
Link DWL-G700AP WAP (91%), Linux 2.6.20
(91%)
7- Service detection is performed on the DMZ
network with the following results: Interesting
ports on 10.19.7.10: Not shown: 999 filtered
ports
$%h!%!$& '!&!$% 
80/tcp open http Thy httpd 0.9.4 (Debian;
GnuTLS
1.0.16; zlib 1.2.2) Service Info: OS: Linux
3. Solution and Analysis 3.1 Firewall testing
results
After performing a testing using nmap tool and
using it to scan and probe the other group
firewall we found the following:
1- The firewall Mac Address which
is FE:FD:00:00:ED 6.
2- the open port, the protocol and the service of
the
dmz.web server:
$%h!%!$& '!
80/tcp open http
3- The number of the up hosts on the dmz
network, which are two hosts.
4- The available open services on the firewall
external interface eth0, Lan interface eth1 and
DMZ interface eth2 which are:
$'(%h!%!$& '!
1 open icmp
5- Identified unfiltered ports on the dmz.web
server
which are:
PORT STATE SERVICE
80/tcp unfiltered http
6- OS detection is performed on the DMZ
network
with the following results:
Device type: WAP|general
purpose|printer|router
Running (JUST GUESSING) : T-Home
embedded
(96%), Linux 2.6.X|2.4.X (94%), FON Linux
2.6.X
(93%), Linksys embedded (92%), D-Link
embedded
(91%), Xerox embedded (90%), Enterasys
embedded (89%), Netgear embedded (89%)

$)
In School of Computer Science,
Telecommunication and Information System [3]
they implemented the new efficient techniques
for firewall testing. In their techniques they avoid
exhaustive and pure random testing
Using their approach, their evaluation study
shows better accuracy and performance than
the random testing. Their approach is also
shown to be robust as it maintains better results
than random sampling even when there is a
small correlation between estimated segments
weight and the probability of error. When
policies of different styles and segmentation
sizes are implemented in their evaluation it
shows that their approach is far better than all
test cases. It is also proved that segmentation
approach has more advantage as rule
interaction.
Currently there research is still in progress.
Studying different segmentation behavior for
several policy styles need further investigation.

*'




In this paper, we have explained importance of
firewall, how we can configure firewall of linux
based operating system using iptables but the
main focus on this paper is on firewall testing. It
gives confidence to the administrator that
firewall is working fine and system is secure.
Before that we described the usage of iptables.
Then we described testing methods. The
technologies we used in our project environment
are UML, MLN and iptables. After applying
method to some of these environments we get
some results.
Our findings were that firewall testing should be
given more planning and resources; because
there are many variables involved in the testing
process and for each variable there are many
values, and for both variables and values there
are many combinations to test.


$


[1] Yong Du and Daniel Hoffman. "PBit-A
Pattern-Based Testing Framework for iptables",
Proceedings of the Second Annual Conference
on Communication Networks and Services
Research (CNSR'04), IEEE (2008),
[2] Linux Home Networking,
"http://www.linuxhomenetworking.com/wiki/index
.php/Quick_HOWTO_:_Ch14_:_Linux_Firewall
s_Using_iptables", (28, April 2009).
[3] Adel El-Atawy, Khaled Ibrahim, Hazem
Hamed, and Ehab Al-Shaer. "Policy
Segmentation for Intelligent Firewall Testing",
Secure Network Protocols, 2005. (NPSec), 1st
IEEE ICNP Workshop on 6 Nov. 2005
Page(s):67 - 72.
[4] The User-mode Linux Kernel Home Page,
"www.user-mode-linux.sourceforge.net", (30
April
2009).
[5] The Manage Large Networks project,
"www.mln.sourceforge.net", (2 May 2009).
[6] Andrew J. Bennieston,"NMAP - A Stealth
Port Scanner", http://nmap.org/bennieston-
tutorial/, (3
May 2009).

Reference: http://www.seminarprojects.com/Thread-firewall-configuration-and-testing-full-
report#ixzz19OBGMDT9c