Apache Web server Complete Guide

Apache Web server
Complete Guide

Dedoimedo

Foreword

Since I cannot be sure you have read my introductory article on
my website (http://www.dedoimedo.com/), here’s an abstract of
what you should expect from this document.

The Web server - Apache - Complete Guide is one of the many
topics covered in the series of books that I’m writing on Linux,
the goal of which is to help any enthusiastic Windows user or a
Linux newbie become a powerful, conﬁdent Linux professional. As
a preview of what you should expect when these books become
published, I have decided to post a single Part on my website.

I am truly convinced that you will thoroughly enjoy this docu-
ment, for it has been written with care and attention to tiniest
details. Every procedure is explained step by step, accompanied
by numerous examples and screenshots.

I hope this will be the best guide on the Apache Web server you
will have ever read.

The only thing that you will miss is the fact that links to other
Parts, covering other material, are not available in this stand-
alone release. However, every procedure required to setup the Web

1

www.dedoimedo.com all rights reserved

server is fully self-contained. You will be able to fully conﬁgure the
Apache server by just using this document as your guide.

Lastly, let’s get one thing straight: you will not become Apache gu-
rus by reading this document. For that matter, I’m not an Apache
guru, either. There are so many aspects to the usability and secu-
rity of the Apache Web server, it is practically impossible to put
them all in a single book.

However, by reading this document, you WILL learn how to use
the Apache Web server on the basic and intermediate level. And
there’s no guide that will explain things as simply and as beautifully
as mine, I guarantee that.

From here on, the sky’s your limit.

2

About

Dedoimedo is a website specializing in step-by-step tutorials in-
tended for human beings. Everything posted on my website is
written in plain, down-to-Earth English, with plenty of screenshot
examples and no steps ever skipped. You won’t easily find tutorials
simpler or friendlier than mine.

Myself, I’m a former physicist, currently living the dream and
working as a Linux Systems Expert, hacking the living daylight
out of the Linux kernel. Few people have the privilege to work in
what is essentially their hobby and passion and truly love it, so I’m
most grateful for the beauty, freedom and infinite possibilities of
the open-source world. I also hold a bunch of certifications of all
kinds, but you can read more about those on my website.

Have fun!

3

Copyright

This document can be used under the following conditions:

If you want to modify or extend any part of this document, please
contact me by email for permission. In any case, you must publicly
credit me for the original work, including a link to my website.

If you wish to mirror either the original article on my website or
just this document, please contact me by email for permission. If
you want to hotlink, please do so with a complimentary explana-
tion, necessary credits and a link to my website.

If you want to use this document for commercial or business pur-
poses, please contact me by email with the details of your endeavor
so we can discuss it.

4

Disclaimer

I am not very fond of disclaimers, but they are a necessary part of
our world. So here we go:

I must emphasize the purpose of this document is for educational
purposes. It is not an oﬃcial document and should not be treated
as such. Furthermore, I cannot take any responsibilities for errors,
inaccuracies or damages resulting from the use of this article (and
its contents).

All of the material in this document has been carefully worded
and prepared by me. However, if for some reason you may feel this
document infringes on copyright or intellectual property of another
work, please contact me with a detailed explanation pointing to a
troublesome part and I will try to sort the problem in the best way
possible.

This tutorial has also been posted as a web article on my website.
For any news, changes or updates, you should always refer ﬁrst to
www.dedoimedo.com.

5

Contents

1 Introduction 11

2 Basic Setup 13
2.1 Verify installation . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Package files . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Main configuration file(s) . . . . . . . . . . . . . . . . . . . . . 15
2.4 Backup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Edit the httpd.conf configuration file . . . . . . . . . . . . . . 16
2.5.1 ServerRoot . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5.2 PidFile . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5.3 ServerName . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5.4 /etc/hosts file . . . . . . . . . . . . . . . . . . . . . . . 19
2.5.5 DocumentRoot . . . . . . . . . . . . . . . . . . . . . . 23
2.5.6 ErrorLog . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.7 Listen . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Create your HTML documents . . . . . . . . . . . . . . . . . . 26
2.7 Start the Web Server . . . . . . . . . . . . . . . . . . . . . . . 28
2.8 Access the web site . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8.1 Local access . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8.2 External access . . . . . . . . . . . . . . . . . . . . . . 32
2.9 Summary of basic setup . . . . . . . . . . . . . . . . . . . . . 36

6


3 Advanced setup 37
3.1 Directory tags . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.1 Order (allow, deny) . . . . . . . . . . . . . . . . . . . . 40
3.1.2 Indexes . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.1.3 DirectoryMatch . . . . . . . . . . . . . . . . . . . . . . 48
3.2 Files tags . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3 Location tags . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.4 Directory, Files and Location . . . . . . . . . . . . . . . . . . 51
3.5 Redirect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.6 Virtual Hosts . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.6.1 Single IP, two websites . . . . . . . . . . . . . . . . . . 57
3.6.2 Two IPs, two websites . . . . . . . . . . . . . . . . . . 63
3.6.3 Other scenarios . . . . . . . . . . . . . . . . . . . . . . 69
3.6.3.1 Different content for intranet and Internet . . 69
3.6.3.2 Different websites on different ports . . . . . . 72
3.7 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.7.1 Module types . . . . . . . . . . . . . . . . . . . . . . . 74
3.8 View installed modules . . . . . . . . . . . . . . . . . . . . . . 74
3.8.1 LoadModule . . . . . . . . . . . . . . . . . . . . . . . . 76
3.8.2 mod_access . . . . . . . . . . . . . . . . . . . . . . . . 78
3.8.3 mod_dir . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.8.4 mod_perl . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.8.5 mod_python . . . . . . . . . . . . . . . . . . . . . . . 79
3.8.6 mod_ssl . . . . . . . . . . . . . . . . . . . . . . . . . . 79

4 .htaccess 80
4.1 Create .htaccess file . . . . . . . . . . . . . . . . . . . . . . . . 83
4.2 Create .htpasswd file . . . . . . . . . . . . . . . . . . . . . . . 83
4.3 Copy .htaccess to restricted directory . . . . . . . . . . . . . . 85
4.4 Configure httpd.conf to allow authentication via .htaccess . . . 85
4.5 Restart server . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

7


4.6 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.7 Other configurations . . . . . . . . . . . . . . . . . . . . . . . 90
4.7.1 Inheritance & performance loss . . . . . . . . . . . . . 90
4.7.2 Disable web access to .htaccess . . . . . . . . . . . . . 91

5 Secure Web server 93
5.1 Encrypted session . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.2 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.4 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.4.1 Main configuration file(s) . . . . . . . . . . . . . . . . . 97
5.4.2 Backup . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.3 Edit the ssl.conf configuration file - part 1 . . . . . . . 98
5.4.3.1 LoadModule . . . . . . . . . . . . . . . . . . 98
5.4.3.2 Listen . . . . . . . . . . . . . . . . . . . . . . 99
5.4.3.3 VirtualHost . . . . . . . . . . . . . . . . . . . 99
5.4.4 Create SSL certificate . . . . . . . . . . . . . . . . . . 101
5.4.4.1 Create Certificate Authority (CA) . . . . . . 102
5.4.4.2 Create server key . . . . . . . . . . . . . . . . 107
5.4.4.3 Create Certificate Signing Request (CSR) . . 108
5.4.4.4 Sign Certificate Signing Request (CSR) with
Certificate Authority (CA) . . . . . . . . . . . 110
5.4.4.5 Verify certificates . . . . . . . . . . . . . . . . 112
5.4.5 Edit ssl.conf configuration file - part 2 . . . . . . . . . 115
5.4.5.1 Server Certificate . . . . . . . . . . . . . . . . 115
5.4.5.2 Server Private Key . . . . . . . . . . . . . . . 116
5.4.5.3 Certificate Authority . . . . . . . . . . . . . . 116
5.4.6 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.4.7 Mini-summary . . . . . . . . . . . . . . . . . . . . . . . 122
5.4.7.1 Names . . . . . . . . . . . . . . . . . . . . . . 122
5.4.7.2 Commands . . . . . . . . . . . . . . . . . . . 123

8


5.4.7.3 Difference between self-signed and CA-signed
certificates . . . . . . . . . . . . . . . . . . . 124
5.4.7.4 Verification . . . . . . . . . . . . . . . . . . . 125
5.4.7.5 File names and locations . . . . . . . . . . . . 125
5.5 Extras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.5.1 Do not use password-protected server keys . . . . . . . 126
5.5.1.1 Create server key without password . . . . . . 126
5.5.2 Submission of CSR to CA . . . . . . . . . . . . . . . . 128
5.5.2.1 Create CSR . . . . . . . . . . . . . . . . . . . 128
5.5.2.2 Send CSR to CA . . . . . . . . . . . . . . . . 129
5.5.2.3 Verify certificate . . . . . . . . . . . . . . . . 129
5.6 General considerations . . . . . . . . . . . . . . . . . . . . . . 130
5.6.1 Use secure server only . . . . . . . . . . . . . . . . . . 130
5.6.2 Use only IP-based virtual hosts . . . . . . . . . . . . . 130
5.6.3 Use server.key as file name for the server key . . . . . . 131

6 Other configurations 132
6.1 Firewall rules . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.1.1 Advanced firewall rules . . . . . . . . . . . . . . . . . . 133
6.1.1.1 Port forwarding . . . . . . . . . . . . . . . . . 134
6.1.1.2 Destination NAT . . . . . . . . . . . . . . . . 135
6.1.1.3 Static NAT . . . . . . . . . . . . . . . . . . . 136
6.2 Enable Web server on startup . . . . . . . . . . . . . . . . . . 139

7 Security 140
7.1 Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.2 Hide your server version . . . . . . . . . . . . . . . . . . . . . 141
7.3 Logs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.4 Permissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.5 Access to root (/) . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.6 AllowOverride . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

9


7.7 Disable public access to .ht files . . . . . . . . . . . . . . . . . 146
7.8 Dynamic content . . . . . . . . . . . . . . . . . . . . . . . . . 146
7.8.1 Disable CGI . . . . . . . . . . . . . . . . . . . . . . . . 146
7.8.2 Disable Server Side Includes (SSI) . . . . . . . . . . . . 147
7.9 Disable unnecessary modules . . . . . . . . . . . . . . . . . . . 147
7.10 Use ModSecurity (mod_security) module . . . . . . . . . . . . 147
7.11 Chroot Jail . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
7.12 Secure web server only . . . . . . . . . . . . . . . . . . . . . . 149
7.12.1 Different DocumentRoot . . . . . . . . . . . . . . . . . 149
7.12.2 Permissions . . . . . . . . . . . . . . . . . . . . . . . . 149
7.12.3 Duration of certificates . . . . . . . . . . . . . . . . . . 149
7.13 Word of caution . . . . . . . . . . . . . . . . . . . . . . . . . . 150

8 Additional resources 151

9 Exercises 152
9.1 Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
9.1.1 Secure Web server & VirtualHost . . . . . . . . . . . . 153
9.1.2 Directory, Files and Locations . . . . . . . . . . . . . . 154
9.1.3 Server functionality, 1 . . . . . . . . . . . . . . . . . . 154
9.1.5 .htaccess . . . . . . . . . . . . . . . . . . . . . . . . . . 156
9.2 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
9.2.1 Secure Web server & VirtualHost . . . . . . . . . . . . 157
9.2.2 Directory, Files and Locations . . . . . . . . . . . . . . 157
9.2.5 .htaccess . . . . . . . . . . . . . . . . . . . . . . . . . . 159

10

Chapter 1

Introduction

A Web server is a server that is responsible for accepting HTTP
requests from web clients and serving them HTTP responses, usu-
ally in the form of web pages containing static (text, images etc)
and dynamic (scripts) content.

The Apache Web server has been the most popular and widely
used Web server for the last decade. It is used by approximately
50% of all websites. Apache is cross-platform, lightweight, robust,
and used in small companies as well as large corporations. Apache
is also free and open-source.

The Apache Web server has almost endless possibilities, due to
its great modularity, which allows it to be integrated with numer-
ous other applications. One of the most popular bundles is the
LAMP Web server application stack, which includes the Apache
Web server alongside MySQL, PHP, Perl, and Python.

The Apache Web server is developed by the Apache Software Foun-
dation. You can read more about Apache on Wikipedia.

11


Being able to configure and secure the Apache Web server is one
of the most important tasks for a (Linux) system administrator.
Almost every company has some sort of a website that advertises it,
including intranet pages that are used by the company’s workers.
The Web interface is used for many tasks beside pure browsing,
including tasks as simple as meal orders and shift rosters, but also
important tasks like administration of databases. In most cases, a
local web server is setup to accommodate these needs.

If you are working for a company that hosts public websites, the
task becomes even more complicated. Web sites are used to serve
content to billions of users daily. Whoever controls this content
- controls the World Wide Web, from news and blogs to financial
transactions. Web servers are hubs of information and power. Mis-
configured or compromised servers can expose a large number of
people to undesired content and potentially incur huge damages to
involved parties.

Running a Web site is much more than opening a port and serving
a few HTML pages. There are tremendous network usability and
security considerations that must continuously be met, evaluated
and improved in order to maintain a safe and effective Web server.

In this Part of the Book, we will learn how to properly setup and
run the Apache Web server, including the secure (HTTPS) server.

12

Chapter 2

Basic Setup

In this chapter, we will setup a Web server that will serve pages
on our internal network. In this chapter, we will perform the most
basic setup with the minimum number of steps required to get
the server running. Later, we will slowly expand, introducing new
features and options.

2.1 Verify installation
First, we’ll verify that Apache is indeed installed:

rpm -q httpd

If you get an empty prompt or a message saying the package is
not installed, you will need to download and install it. If the shell
displays the package name and version, you’re good to go.

13


2.2 Package ﬁles
Rule no. 1: don’t panic! The list before you might seem intimi-
dating at the moment, but that is simply because you are not yet
familiar with Apache. But don’t worry. For now, treat the list as
a reference only. At this stage, you don’t need to know anything
or remember anything. We will slowly cover everything, step by
step.

Now, let us overview the location and purpose of the ﬁles used by
the Apache server. Please note that the list is partial and includes
only the most important entries. We will slowly expand this list
as we go through the Part.

14


File name Description
/usr/sbin/httpd server binary
/etc/httpd directory containing server
configuration files
/etc/httpd/conf directory containing main
configuration files
/etc/httpd/conf.d directory containing
configuration files for
individually packaged
modules, like ssl, php, perl etc
/etc/httpd/logs symbolic link to
/var/log/httpd
/etc/httpd/modules symbolic link to
/usr/lib/httpd/modules
/etc/httpd/run symbolic link to /var/run
/usr/lib/httpd/modules server modules
/var/log/httpd server log
/var/run runtime variables
/var/run/httpd.pid server process ID
/var/www/html public html files

We will discuss the main configuration file shortly.

2.3 Main configuration file(s)
The main configuration file for the Apache Web server is:

/etc/httpd/conf/httpd.conf

15


This file is well commented and self explanatory. It contains quite
a large number of settings, but we’ll concentrate on just the few
necessary to setup the server.

2.4 Backup
This is one of the most important things to remember. Always
retain the copy of the original file so you can easily revert to the
default. At the very least, do NOT delete default lines; instead,
just comment them out so you’ll be able to see what the original
settings read and refer to them.

cp /etc/httpd/conf/httpd.conf→
/etc/httpd/conf/httpd.conf-default

2.5 Edit the httpd.conf configuration file
Let’s open this file in vi text editor and review the most important
entries. The file has many options - but we need only a few. In fact,
you will need to change just a single line to create your server and
get it running. However, you should be familiar with some other
settings before launching the server.

vi /etc/httpd/conf/httpd.conf

This is what the file looks like - at least the beginning of it:

16


Let’s go over the most important entries you should remember for
now.

2.5.1 ServerRoot
ServerRoot is the path to the server’s configuration, error and log
files. It is possible to change this path, provided all the necessary
files are copied to the new location accordingly. We will later
review this concept as a part of the security measure known as the
Chroot Jail, but more about that later (). The default location is
/etc/httpd.

17


As you can see, the file is rich with easy-to-understand comments.
Apropos the comments, please note that you should not place a
trailing slash at the end of the specified path.

2.5.2 PidFile
PidFile is the process identification number for the httpd. This
process number is important, because Apache spawns numerous
child processes when running to accommodate the web traffic. It
allows you to monitor and manipulate your server processes. See
image above.

2.5.3 ServerName
This is the one setting you will have to change to get your server
running. This is where you declare the name of your website.

18


I will use www.ninja.com - just a random name with no association
whatsoever to the real site bearing this name.

The generous comments in the file remind us that if we do not have
a registered DNS name, we should use an IP address. One, we’ll
discuss registered DNS names later. Two, we’re going to use the
hosts file to demonstrate the address-to-name translation.

2.5.4 /etc/hosts file
As you already know, the hosts file allows easy matching of names
to IP addresses, . In general, using the hosts file is a good way
of testing your IP-to-name (or vice versa) configurations before
committing these changes into a production environment.

19


First, with no new entries added to the hosts ﬁle, typing
www.ninja.com in the address bar of a web browser takes us to
the site itself (on the Internet).

Now, we shall edit the ﬁle and add an entry, pointing www.ninja.com
to a local IP address.

vi /etc/hosts

20


After saving the hosts ﬁle, we can no longer see the Internet site.
Furthermore, we don’t get any fancy results from our own Web
server, because it is not running yet.

21


22


2.5.5 DocumentRoot
DocumentRoot tells you where your web documents (html files, im-
ages etc) should be located. It is possible to reference files in other
directories using aliases and symbolic links. The default directory
is /var/www/html.

2.5.6 ErrorLog
ErrorLog tells you where the log containing all server errors is
located. This file is critical for debugging and solving server mis-
configuration problems and for proper traffic shaping. By default,
all messages with the value of warning (warn) and higher will be
logged. This is described in the LogLevel directive just below.

23


The default location is logs/error_log. Please note that this is rela-
tive to the ServerRoot. Therefore, our log ﬁle is
/etc/httpd/logs/error_log. However, let us not forget that
/etc/httpd/logs is a symbolic link to /var/log/httpd. Thus, ﬁnally,
the actual error log is /var/log/httpd/error_log.

24


2.5.7 Listen
The Listen command tells the Web server what ports to use for
incoming connections. By default, port 80 is used, although any
one or several can be used. The accepted conventions calls for
using port 80 for non-secure web communications (without any
encryption of traffic). Secure web communications are normally
handled on port 443.

We’ll talk about the Secure Web server later.

That’s it. These are all the settings you need to know for now and
tamper with in order to successfully launch the Web server. Save
the configuration file (Esc then :x in vi text editor).

25


2.6 Create your HTML documents
Now, just to make things more interesting, we shall create a num-
ber of files and place them in the DocumentRoot directory
(/var/www/html), including a simple index.html file. Here’s the
source of our index.html file (the two echoes are used to make the
output easier to read):

26


And here is the preview of ﬁles we have in the html directory:

Now that we know what we have, it’s time to power up the server.

27


2.7 Start the Web Server
Start the httpd service:

service httpd start

If everything worked out ﬁne, the web server should start without
any errors and you should see the following image:

Still, it does not hurt to check the status of the service or verify
its process ID:

28


There are 9 processes running for Apache. This may be confusing,
but there’s a very simple explanation for this. In the httpd.conf
file, you will find a directive called StartServers. This directive tells
the Web server how many server processes to launch on startup.
The default setting is 8 server processes.

Once started, the Web server dynamically kills and creates pro-
cesses based on the traffic load, with the number of server processes
fluctuating between MinSpareServers and MaxSpareServers. So
far, everything figures out just nicely. Now, let’s make another
check.

The Apache Web server, if configured to listen on port 80 (or any
“secure” port below 1024) must be started as root. Otherwise,
it can also be started by regular (non-root) users. As a security

29


precaution, the server processes spawned by Apache run as user
apache, which belongs to the group apache. Indeed, we can easily
verify that:

You can change these settings in the httpd.conf file, as well.

2.8 Access the web site
2.8.1 Local access
Now, let’s access our homepage.

Open a web browser and type www.ninja.com in the address bar.
Earlier, we were unable to access it, even though we have specified
the entry for our website in the hosts file. This was because the

30


server was not running. But now, www.ninja.com resolves to our
custom webpage.

Our server works.

Alternatively, we could have simply accessed it by typing localhost
in the web browser address bar.

31


2.8.2 External access
Accessing the web server locally is not the most challenging thing
to do. Let’s access it from other machines.

32


Here’s our webpage, seen in another CentOS machine, belonging
to the same subnet:

33


Here’s our webpage, seen in the Opera browser running on a Win-
dows machine:

As you might expect, using www.ninja.com on the other machines
does not yield the desired result. This is because there is nothing
telling these machines to match the name to the IP address of
our server (192.168.1.128). On the server itself, we overcame the
problem using the hosts ﬁle.

Theoretically, we could do the same thing on every host on our
network, but this is slightly impractical and cumbersome. How-
ever, this will not solve the problem of accessibility from hosts that
we have no control of, outside our local network. To overcome this
monumental problem, we’ll use name resolution by conﬁguring and

34


running a Domain Name System (DNS) server. This material is
covered in great detail in the next Part.

Meanwhile, everything works as we’ve expected. Soon, we will go
over some advanced conﬁgurations.

35


2.9 Summary of basic setup
To make things simple and clear, here’s an overview of the steps
you will have to take to setup and launch Apache:

• Verify installation of the Apache RPM.

• Backup the /etc/httpd/conf/httpd.conf main configuration file.

• Open it in the vi text editor and review the options listed therein.

• Setup the DocumentRoot directive (default /var/www/html).

• Setup the ServerName directive (for example, www.ninja.com).

• Optionally setup other directives (like ServerRoot, ErrorLog, Listen
etc).

• Configure the /etc/hosts file so that you can access the website by
name.

• Create a sample HTML file and place it in the DocumentRoot directory.

• Start the httpd server.

• Test the setup by accessing the web site.

36

Chapter 3

Advanced setup

The httpd.conf file can be extensively customized using a range
of directives. We have studied a few and will now review several
more. Please note that it is impossible to list every single directive
here. Nevertheless, we will go over some of the more useful and
practical directives, which will greatly enhance the usability (and
also the security) of your web server.

3.1 Directory tags
Directory tags allow you to specify the configurations separately
for each directory serving the web pages. If you are familiar with
HTML and CSS, then using <div> containers might be the sim-
plest analogy. This allows you to serve content to specific IP ranges
while denying other ranges, limit access to certain files, set the be-
havior of pages contained in these directories, and more.

37


Just about any directory can be listed, although it is not necessary.
The most sensible solution is to setup very restrictive parameters
to the root (/) directory and custom, desired parameters to direc-
tories inside DocumentRoot.

Directory tags take the following form (again this is very analogous
to HTML <div> tags):

• <Directory directory_path> tag begins a block.

• Next, follows a series of options deﬁning what users accessing web pages
located in this directory can do.

• </Directory> tag closes the block.

38


Here’s a sample block, showing the default settings applied to the
root (/) directory:

<Directory />
Options FollowSymLinks
AllowOverride none
</Directory>

Let’s try to understand what we have here:

<Directory />

This declares the block for the root (/) directory and all sub-
directories.

Options FollowSymLinks

The Options directive declares which server features are valid for
the specified directory; FollowSymLinks is one of the possible op-
tions - it allows webpages to use symbolic links to point to files
located anywhere on the root (/) directory. Please note this is not
the best configuration from the security point of view; however, it
does demonstrate the functionality of the Directory tags. We will
discuss the server security measures later in the Part.

AllowOverride none

The AllowOverride directive governs the behavior of .htaccess files
(more about them later). It tells whether the restrictions imposed
by the Options can be overridden by specific settings inside the
.htaccess files. The default behavior is set to none and should

39


remain that way. This will prevent security breaches or nuisances
due to misconﬁguration.

</Directory>

This tag closes the block.

3.1.1 Order (allow, deny)
Allow and Deny directives govern the access to the directory de-
clared (via the Directory tags). The Order directive speciﬁes how
the allow and deny directives are treated. The Order of allow,
deny can be looked upon as default-allow or blacklist; only “bad”
hosts or IPs are disallowed. The Order of deny, allow can be
looked upon as default-deny or whitelist; only “good” hosts or IPs
are allowed.

Possible declaration of allowed or denied clients can be via host
name, domain name, IP address, partial IP address, and more.

Here, we’ll restrict access to the directory (or rather, the server)
by denying access from all - and only permitting access from a
single IP address, that of another machine on the LAN (in this
case, 192.168.1.129).

40


Let’s review the changes to the httpd.conf file:

• I have commented out the original parameters, which allowed access
from all hosts (or IPs).

• I have changed the order of allow, deny directives. Again, this is im-
portant, because the order defines the precedence of the rules. Thus,
first, we’ll deny everyone (this can be called default deny policy, so to
speak) and then permit only specific hosts (or IPs). If the Order were
reversed (allow, deny rather than deny, allow), no one would be able
to access the server. This is critically important to remember when
implementing allow, deny policies.

41


The changes will only take effect after the Web server is restarted
or the configuration file reloaded. This can be achieved by running
either:

service httpd restart

Or:

service httpd reload

After httpd reads the new configuration file, the changes will take
effect. Now, let’s try to access the server from the Windows ma-
chine.

42


As you can see, we are denied access. But accessing from the
CentOS client with the IP of 192.168.1.129 works ﬁne.

3.1.2 Indexes
The Indexes directive tells the server whether to display the di-
rectory listing when asked. The behavior of this directive depends
on another directive - the DirectoryIndex. The DirectoryIndex di-
rective tells the server the name of the default page that it should
serve when a user requests the listing of a directory.

This is the typical everyday scenario. Users are trying to access
webpages by simply typing their names, without typing the ex-
act homepage (like index.html, index.php etc). Various ﬁle names

43


specified under the DirectoryIndex are looked for and the first one
found is presented to the user. If no file is found, the listing of the
directory is then generated by the server.

This is something you may want to avoid, especially if there are files
you do not wish your users to see. However, if the Options Indexes
directives are used, then directory listings will be generated.

One solution is to place a dummy index.html file in every directory,
but this is cumbersome. The more elegant approach is to disable
the listing globally (remove Indexes from the Options directive
under DocumentRoot) and then allow per-directory listing when
you see fit.

The default configuration in the httpd.conf file specifies Options
Indexes for the Directory tags of the default DocumentRoot
(/var/www/html). We will change that.

First, we will remove Indexes from the Options line for our Docu-
mentRoot. Then, we will create two directories, called index_allow
and index_deny, where only the first will have the Options In-
dexes specified. Both of these directories will contain some random
files.

44


This is the new conﬁguration ﬁle. Save it, then restart httpd. Now,
if we request the directory listing for each one from our clients, we’ll
get the following results:

45


index_allow

46


index_deny

47


3.1.3 DirectoryMatch
The directives enclosed in the Directory tags will be indiscrimi-
nately applied to all sub-directories. If you require a more fine-
tuned approach for several similar sub-directories, you will have
to use the DirectoryMatch tags. The main difference is that the
DirectoryMatch tags allow the use of regular expressions, allowing
you to match several sub-directories inside a single rule.

Again, for those familiar with HTML / CSS and the use of classes
and ids, the idea is very much similar.

3.2 Files tags
The Files tags are very similar to the Directory tags. The ma-
jor difference is that while the Directory tags govern the scope of
permissions (or restrictions) of the enclosed directives by directory
name, the Files tags do the same on the file name level. In other
words, the Files tags can be used to configure the behavior of a
single file - or a set of files that match a regular expression.

Here’s an example, showing the restrictions applied to .htaccess
and .htpasswd files, the files usually used in restricting access to
certain directories (and/or files) by requiring users to authenticate
before viewing the content:

48


We will review this particular example later in this Part.

In the above example, we’ve seen the use of regular expressions
to allow multiple files to be covered by a single rule. However, a
comparable directive, more suitable for handling multiple files and
complex regular expressions is the FilesMatch directive.

3.3 Location tags
Again, the Location tags are quite similar to the two mentioned
above. The major difference is that the Location tags are used to
limit the scope of enclosed directives by URLs.

49


In other words, the Directory and Files tags should be used to con-
trol content that resides on the system (like various ﬁles and im-
ages, within their sub-directories), while the Location tags should
be used to control content that is located outside the system, like
databases, for instance.

Below, we can see a commented example included in the httpd.conf
ﬁle. If enabled, this block would allow you to access your server
statistics, but only if you connected from the server itself.

50


Here’s an example (please disregard the actual URL):

Again, for complex regular expressions, you should use the Loca-
tionMatch directive.

3.4 Directory, Files and Location
The Directory, Files and Location tags all perform a similar func-
tion: they categorize what restrictions are placed on content en-
closed by each one. At first glance, there seems to be very little
difference between them. However, just like the order of allow and
deny directives is critical, so is the correct use of these tags.

The configuration sections must be placed in a very particular order
to make sure they behave as intended. The order of precedence of

51


their execution by the server means that a misplaced section could
compromise the security of the server - or not get executed at all.

For more details, please refer to the following Apache documenta-
tion page: Conﬁguration Sections - Apache HTTP Server.

3.5 Redirect
The Redirect setting allows you to map an old webpage to a new
URL. This could be the case if you changed domain, for example,
or moved around a lot of ﬁles, renaming and deleting them. To
demonstrate the directive, we’ll map our server to point to my
own site.

52


Save the ﬁle, restart the server.

53


3.6 Virtual Hosts
Virtual Hosts is an important, powerful feature that allows you to
run several websites from a single computer. Virtual Hosts can
be IP-based or named-based, offering a high level of customization
(and flexibility).

Virtual Hosts can use almost any option normally used in the
httpd.conf file. To make you better understand this, you can treat
Virtual Hosts as individual customized httpd.conf files nested in-
side the main httpd.conf file.

54


Let’s review a sample Virtual Host:

<VirtualHost *:80>
DocumentRoot /var/www/html/ninja-father
ServerName www.ninja-father.com
# other directives
</VirtualHost>

What do we have here?

<VirtualHost *:80>

This declares the name or the IP address of the site (server) that
should be served using the directives inside the VirtualHost block
on port 80. If no port number is used, the default one speciﬁed
under the Listen option is used. The default port is 80 (standard
convention). Asterisk (*) can be replaced with any name (for exam-
ple, www.ninja.com) or IP address (for example, 192.168.1.128),
depending on your needs and requirements. Let see several simple
examples:

• <VirtualHost 192.168.1.128:80> will apply the directives listed in the
block below to all incoming connections aimed at 192.168.1.128 on port
80.

• <VirtualHost 192.168.1.128> will apply the directives listed in the
block below to all incoming connections aimed at 192.168.1.128 on the
default port (as speciﬁed in the Listen directive). If this port is 80,
then this option is identical to the one above.

• <VirtualHost planck.matter.com> will apply the directives listed in the
block below to all incoming connections aimed at planck.matter.com on

55


the default port (which can be 80, 8080 or any other).

• <VirtualHost ninja.com:8777> will apply the directives listed in the
block below to all incoming connections aimed at our site ninja.com on
port 8777. This port must be specified under the Listen directive.

DocumentRoot /var/www/html/ninja-father

This declares the directory where you should place all files that you
wish served when the VirtualHost is invoked (matching names or
IPs and the port).

ServerName www.ninja-father.com

This is the name of the server. In other words, this is the address
people will type in the web browser address name in order to get
to your site. In order to successfully resolve this name to the IP
address of the Web Server, we will need to use /etc/hosts file like
before or setup a DNS Server (later).

# other directives

This is just a comment specifying many other options can be used,
including those we have not yet reviewed here.

OK, now that we know what we’re dealing with, let’s create and
test several scenarios.

56


3.6.1 Single IP, two websites
This is one of the most common setups. We will create two websites
- www.ninja-father.com and www.ninja-son.com. Both will reside
on our server, which has the IP address 192.168.1.128. In order to
make them both accessible to the world, we will create two Virtu-
alHost blocks and declare their DocumentRoot and ServerName
separately.

Here’s what we need to do:

• Create directories inside /var/www/html called ninja-father and ninja-
son.

• Create simple index.html ﬁles for each.

• Edit httpd.conf and create our two VirtualHost blocks.

57


Here’s what the httpd.conf looks like:

Now, for the sake of convenience, we will also use the /etc/hosts
file to allow name resolution to work. It is also imperative in our
case, because using the IP address would always point to the first
VirtualHost listed in the httpd.conf file.

58


Please note that specifying an IP address in two different lines is wrong.
The hosts file will always use only the first entry. You should list all
names for a specific IP in a single line.

59


For example, this is incorrect (although it would work in our case):

Don’t mind the commented lines, they are used for other configu-
rations: the first, our standard website; the second, for yet another
VirtualHost scenario, which we will discuss soon.

Now, we shall save the files (both httpd.conf and /etc/hosts) and
restart httpd. Then, using Firefox, we will try to access each one.

60


www.ninja-father.com

61


www.ninja-son.com

It works like magic. Best of all, the user has no idea that these two
sites reside on the same machine.

62


3.6.2 Two IPs, two websites
This is another common scenario. You can assign a different IP
to each website, avoid possible resolution mixups and simplifying
your setup. However, this requires that you either use more than
a single network adapter or create virtual adapters. If you have,
let’s say 14 websites, having 14 physical network devices plugged
into your machines is not the best idea. Using virtual adapters is
the most sensible choice here.

We already have our two websites ready. We just need to create a
virtual network card and then change the httpd.conf file to reflect
the changes.

First, we will create a virtual adapter (eth0:1) with the IP address
of 192.168.1.200.

63


64


Then, we’ll edit the httpd.conf ﬁle:

65


Lastly, we’ll edit the /etc/hosts ﬁle:

After restarting the server, we’ll be able to get to our two sites
easily. Again, the change is completely transparent to the user.

66


www.ninja-father.com

67


www.ninja-son.com

Excellent.

Please note that the conﬁguration of the virtual network adapter is
temporary. You will have to create a network script to preserve the
change between reboots. This setup has covered extensively in Part
?: Networking - sub-part 2: Basic and intermediate conﬁgurations
().

68


3.6.3 Other scenarios
Basically, the above two scenarios cover pretty much everything.
Once you get the hang of VirtualHost setting, creating any which
setup becomes a simple matter. Nevertheless, for the sake of clar-
ity, I will demonstrate several more typical scenarios in the exam-
ples below, including some features not mentioned yet in this Part
of the Book.

3.6.3.1 Different content for intranet and Internet

In practice, this scenario is very similar to having 2 different IPs
serving two different websites, except that you will use one website
but serve different parts of it to different customers.

Let’s assume you wish to achieve the following:

• Allow users on the local network access to all content, but deny some
to users on the Internet.

• Allow users on the local network to list directory index, but deny this
feature to the Internet users.

• Display a different home page to local users and external customers.

• Allow certain custom scripts to be available only to external customers.

69


Here’s what a sample conﬁguration in httpd.conf ﬁle would look
like:

AddHandler cgi-script .cgi

NameVirtualHost 172.16.1.1:80
<VirtualHost 172.16.1.1:80>
DocumentRoot /www/intranet
ServerName www.our-company.com
<Directory /www/intranet>
Option Indexes FollowSymLinks
</Directory>
</VirtualHost>

<VirtualHost 211.211.211.211:80>
DocumentRoot /www/web
ServerName www.our-company.com
<Directory /www/web>
Options +ExecCGI FollowSymLinks
</Directory>
</VirtualHost>

70


This examples introduces a number of concepts we have not yet
seen, so let’s briefly review them:

NameVirtualHost

This directory allows you to map named-based incoming connec-
tions to specific IP addresses. You might ask yourselves why you
need this, when we have seen perfectly good examples before, with-
out this feature. Well, the answer is: if somehow a named-based
request gets “lost” (due to DNS configuration, firewall rules or
similar), it might not match any of the VirtualHost blocks. In
that case, the default settings configured in httpd.conf will be ap-
plied to this request, which could be contrary to your needs. Using
NameVirtualHost forces all incoming connections to a certain IP
address to point to a certain VirtualHost block. This request will
also never fall back to the main configuration, allowing you a com-
plete modularity in your setup.

Thus, in our example, all requests to the internal IP address will
go the VirtualHost with this IP address declared. We can also
see that the users will be able to view directory listings and follow
symbolic links.

+ExecCGI

We see this directive listed under Options in the second block,
which refers to the Internet customers. All incoming connections
on the external IP will go to the VirtualHost with this IP declared.
We can see the users won’t be able to demand directory listings,
but they will be able to follow symbolic links - and execute .cgi
scripts located in this directory.

71


The ExecCGI directive tells the server to allow server-side script-
ing in the specified directory. The plus (+) signs signifies this
Option is used in addition to all those Options already specified
for the root directory. Similarly, the minus (-) sign can remove
some of the privileges, compared to the Options already specified
for the root directory.

However, alone, this directive is insufficient to allow scripting in
this directory.

AddHandler cgi-script.cgi

In order to enable .cgi scripts to work outside the default script
directory, a directive must be added to the httpd.conf configuration
file. Indeed, this is the first line of our sample code - AddHandler
cgi-script .cgi - it allows scripts in non-default directories to be
executed, by using the +ExecCGI option, as we’ve done before.

The Apache Web server has many other options and features. You
are welcome to try them all, using this Part of the Book as the
foundation for expanding your knowledge. For more information,
please refer to:

• Apache HTTP Server Version 2.o Documentation

• RedHat Enterprise Linux 4: Reference Guide, Chapter 10: Apache
HTTP Server

3.6.3.2 Different websites on different ports

We’ve already discussed this before. Let’s say you have a single
IP address with multiple websites served. Using the hosts file or

72


DNS resolution is a possibility, but this might not always work.
Conﬁguring the Web server to listen on several ports for incoming
connections and then using NameVirtualHost feature to force the
connections to speciﬁc VirtualHost blocks will force the server to
behave as you desire.

Here’s an example:

Listen 192.168.1.128:80
Listen 192.168.1.128:9021

<VirtualHost 192.168.1.128:80>
DocumentRoot /www/white-socks
ServerName www.white-socks.com
</VirtualHost>

<VirtualHost 192.168.1.128:9021>
DocumentRoot /www/black-socks
ServerName www.black-socks.com
</VirtualHost>

For more examples, please refer to: VirtualHost Examples - Apache
HTTP Server.

73


3.7 Modules
Modules are extensions that enhance the basic functionality of the
Web server. The modules reflect the growth of the Web and the
inclusion of dynamic content into the web pages. The static HTML
can provide only so much functionality. In fact, many of the options
we have seen and used above are provided by different modules.
For example, the Order directive is provided by the mod_access
module.

3.7.1 Module types
There are two types of modules:

• Built-in modules, which are compiled into Apache and will load with
the server any time it is started. Their functionality cannot be removed
without recompiling the package. These modules are also known as
static.

• Loadable modules, which can be loaded on and off as required. These
are the shared modules.

3.8 View installed modules
You can always list the modules currently used by the server.
The command below will display only the modules compiled into
Apache.

httpd -l

74


75


This command will list all modules, both static and shared:

httpd -M

There is a wide range of modules available. We will review a num-
ber of more common ones. Please note that the list below is only
partial and just brieﬂy introduces the range of available modules.

3.8.1 LoadModule
Shared modules are called by the Web server using the LoadModule
directive in the httpd.conf ﬁle. If you do not wish to use a certain
module, simply comment its line. However, you must remember
this will remove the functionality that the module provides.

76


These modules are referenced by a symbolic link in the /etc/httpd/
directory, pointing to /usr/lib/httpd/modules.

77


Let us go over some of the more interesting modules, just a sam-
pling.

3.8.2 mod_access
This module provides access control based on client host name, IP
address, or other characteristics of the client request.

3.8.3 mod_dir
This modules provides interface for redirects and serving directory
indexes. We have reviewed quite a bit of its functionality in the
previous sections.

78


3.8.4 mod_perl
This module allows dynamic content produced by Perl scripts to
be served to incoming requests without using the Perl interpreter
every time, reducing overhead and system load. This is done by
embedding a Perl interpreter into the Apache server. The module
can also emulate a CGI environment, allowing the reuse of Com-
mon Gateway Interface (CGI) scripts without any changes to the
setup.

3.8.5 mod_python
mod_python allows integration of the Python programming lan-
guage into the Apache server. It is intended to replace CGI as
a method of executing Python scripts on a web server. It oﬀers
much faster execution and allows data to be maintained over mul-
tiple sessions.

3.8.6 mod_ssl
This module provides an interface to the OpenSSL library, allowing
the use of Secure Socket Layer (SSL) and Transport Layer Security
(TSL) secure communication protocols. This allows you to run a
Web server that will run encrypted sessions with clients, allowing
a safe exchange of potentially sensitive data. We will discuss this
module again when we setup a secure Web server (7.12).

For a detailed list of available modules and their functionality,
please refer to: Apache HTTP Server Module Index.

79

Chapter 4

.htaccess

.htaccess stands for hypertext access. This is the default name of
the Apache directory-level configuration file. This file can be used
to create security restrictions for particular directories. One of the
most common uses is to require user authentication in order to
serve certain web pages.

Before we setup .htaccess, there are some things you should re-
member:
• .htaccess is not a replacement for a carefully laid out security plan. You
should use the httpd.conf file to place restrictions on your server. Only
then should you use .htaccess, to further restrict the already allowed
users.

• Do not ever use .htaccess to handle secure or privileged content, like
user data.

• .htaccess file is loaded every time a webpage is requested, incurring a
performance loss.

• Using this file grants individual users an ability to make security modi-
fications to your site, creating possible risks if not properly configured.

80


On the other hand, using .htaccess is useful if you run a multi-
user hosting plan. These users do not have root access to the main
configuration file and their only way of “shaping” traffic is by using
the .htaccess file. In general, the use of the .htaccess file should
be limited to non-root users only.

Before we can setup access-protected pages, we need to briefly
overview the layout and syntax of the .htaccess files. Let’s examine
what a typical .htaccess file looks like. Then, we will combine it
with our web content.

AuthType Basic
AuthName “Restricted web page”
AuthUserFile “/etc/httpd/conf/.htpasswd
require valid-user

AuthType Basic

This line defines the type of authentication. Basic means there is
no encryption and the password hash is sent as clear text. This is
one of the major reasons why .htaccess cannot be considered for
protection of confidential user data.

AuthName "Restricted web page"

When someone tries to access an .htaccess-protected page, a user-
name & password window will pop in the web browser. This win-
dow will bear a title - this is the AuthName. It can be anything
you like.

AuthUserFile /etc/httpd/conf/.htpasswd

81


This line defines the path to a file where user credentials are stored.
This file does not exist, but we will create it soon.

require valid-user

This line indicates only successful authentication attempts will re-
sult in the loading of the page.

Now that we know what we’re about, we will:

• Create an .htaccess file similar to the one above.

• Create the .htpasswd file containing usernames & password necessary
for the authentication.

• Place .htaccess in the directory we wish users to validate before access-
ing the content.

• Tell httpd to allow user authentication via .htaccess files.

• Restart the server.

• Test the results.

82


4.1 Create .htaccess file

4.2 Create .htpasswd file
First, we will access the directory where we intend to place the
file - /etc/httpd/conf. It can be any directory, but it must be
outside the DocumentRoot, so it so not viewable by your clients.
Furthermore, only the root should be able to modify this file.

Make sure only root can modify the .htpasswd file! It should have
permissions set to 0644.

83


Users and passwords are added to the file by running the htpasswd
command.

htpasswd -c .htpasswd username

The name of the authentication file can be anything. You may
consider changing it to something else.

After you have finished adding the usernames (there can be one
or more), you can see the contents of the .htpasswd file. The
passwords are encrypted.

84


4.3 Copy .htaccess to restricted directory
We will place the .htaccess file in our DocumentRoot. To make
things interesting, we will also change the site and the homepage
somewhat. Instead of ninja.com, we will serve ourserver.com.
This is the site we will use to configure a DNS server in the next
Part ().

4.4 Configure httpd.conf to allow authentica-
tion via .htaccess
By default, .htaccess files are given no control whatsoever. This
is accomplished by the AllowOverride directive. This directive

85


specifies what the .htaccess files can do - in addition and contrary
to main configuration settings. Please note that this could pose a
security risk. Badly configured .htaccess files can compromise the
security of your system.

We will allow .htaccess to authenticate users. We will replace the
original AllowOverride none to AllowOverride AuthConfig.

4.5 Restart server


86


4.6 Test setup
This is the webpage, seen without any restrictions.

87


Now, after restarting the server, we will be asked for authentication
credentials.

88


If we succeed, we will reach the webpage, like before.

89


If we enter the wrong username & password - or none, we will be
rejected. You can customize the “reject” page, if you like.

4.7 Other configurations
While this pretty much covers the basic setup of .htaccess files,
there are several more things you should remember.

4.7.1 Inheritance & performance loss
Please remember that the .htaccess restrictions are inherited by all
sub-directories that exist in the directory you have placed the file.
This means that whenever one of your clients tries to access a page
in one of the sub-directories, the server will have to make a recur-
sive search up the directory tree until it finds the file. Furthermore,

90


even if it does find the file, the server will have to check up every
directory up the tree to create a complete set of restrictions.

4.7.2 Disable web access to .htaccess
By default, Apache prevents any file beginning with letters .ht
to be visible through the web browser. This is a minor security
consideration, which allows you to keep your .htaccess files safe
from prying eyes, even though they are located in world-readable
location (DocumentRoot directories and sub-directories).

This behavior is governed by the combination of the AccessFile-
Name and Files directives. We have seen this example earlier when
we review the Files tag; now, we can see them in practical use.

91


You can also setup other types of files - or just specific files - from
being accessible - or accessible only to certain hosts.

Indeed, if we try to reach .htaccess through the web browser, we
will be denied access.

92

Chapter 5

Secure Web server

Running a secure Web server is something you should consider if
the daily use of your websites will include an exchange of confi-
dential, private information from your users. Regular Web servers
send and receive traffic in unencrypted form. Unfortunately, this
makes them vulnerable to man-in-the-middle attacks, where a po-
tential attacker could use sniffer tools to log packets en route from
clients to the server and derive sensitive information from them.
This mode of security is completely unacceptable for websites that
must deal in personal data, like bank accounts, medical or financial
records, or others.

The secure Web server eliminates this threat by offering two key
advantages:
• It allows users to verify the identity of the server.

• It allows users to conduct safe transactions with your server by en-
crypting the authentication and the session.

To achieve this, the Apache Web server uses secure communication
protocols like the Secure Socket Layer (SSL) or the Transport Layer

93


Security (TLS) to protect the flow of data.

5.1 Encrypted session
Before we setup a secure server, we should first understand how
encrypted communication between the server and the client is con-
ducted. Let us outline the details of a typical secure session:

• A client tries to connect to port 443 on the secure Web server.

• The client sends a list of available encryption methods it supports; if
the client cannot support encryption, for instance very old browsers,
the connection attempt will be unsuccessful. Modern browsers support
both SSL and TLS without any problems.

• The server will choose the strongest available encryption method that
both sides can support.

• The server will then send back to the client its certificate and the pub-
lic encryption key. The certificate is a sort of an ID, telling the client
important information about the server. To make this information
credible, the certificate must be signed by a reputable Certificate Au-
thority (CA), like EquiFax, Thawte or others. The public key will be
used by the client to generate its own encryption hash should it choose
to accept the server’s certificate.

• The client receives the certificate. In most browsers, the certificate is
first compared to an existing list of authorities. If the digital signature
matches, the certificate will be accepted. If no match is found for the
certificate, the browser might use the Online Certificate Status Protocol
(OCSP) to connect to CAs in real time in an attempt to verify the
certificate. Generally, the use of OCSP is not enabled by default in
most browsers, in order to speed up the authentication process. If no

94


match is found still, the client will be issued a warning by the browser,
informing it that the certificate could not be verified. The user now
must decide whether he/she can take the risk and accept the certificate.
In addition to being self-signed (i.e. no CA signature), the typical
issues arising with certificate prompts include a mismatch between the
site you are trying to access and the one registered in the certificate,
dubious credentials or an expired certificate.

• Regardless of what may occur, if the client accepts the connection, it
will send back a hash encrypted with the server’s public key. This hash
will be used to encrypt all communication between the server and the
client throughout the session. Only the client will be able to decrypt
the communications - or rather, anyone who possesses the private key.
But if the client side is fairly secure and the server’s certificate is valid,
the communication is safe.

5.2 Requirements
We have already mentioned that the client must support some sort
of encryption to able to establish secure connections to a server.
On the server end, the server must also support the secure com-
munication protocols. The Apache Web server uses the mod_ssl
module, which provides an interface to the OpenSSL library, al-
lowing the use of SSL and TLS.

By default, most distributions today ship with the OpenSSL library
installed and the Apache server compiled against the mod_ssl
module. If your distro does not include either one or both, you
will have to obtain them before you can use a secure Web server.

95


You can check if you have the OpenSSL library installed:

rpm -q openssl

And to certify if Apache uses mod_ssl, you should look for it in
the /etc/httpd/modules directory.

5.3 Limitations
On one hand, the secure Web server oﬀers veriﬁcation of the server’s
identity and safe transactions. On the other hand, it is slower than
the regular server. Therefore, you should take into consideration
the performance loss stemming from the use of encryption. You
should not use the secure Web server for regular daily content that
does not include any exchange of personal information.

96


5.4 Setup
5.4.1 Main configuration file(s)
The main configuration file for the secure Apache Web server is:

/etc/httpd/conf.d/ssl.conf

This file is very similar to httpd.conf, except that it includes a
number of special directives. But the principle remains the same.
Basically, the configuration file contains a VirtualHost block, where
all secure Web server directives should be listed. We will edit this
block to suit our needs.

97


5.4.2 Backup
We will first backup the file before making any changes.

cp /etc/httpd/conf.d/ssl.conf →
/etc/httpd/conf.d/ssl.conf-backup

5.4.3 Edit the ssl.conf configuration file - part 1
Again, we need to make a number of changes to get our server
to work. However, before we can fully edit all of the necessary
options, we will have to digitally sign our server. This includes
creating the public key and signing it with a certificate from a
known, reputable CA. However, since we do not have a certificate,
it costs money and the process takes time, for the purpose of this
exercise, we will create our own CA and use it to sign our server.

But first, let us review the most important directives that we need
to get our server started. The procedure is identical to what we
have done earlier.

5.4.3.1 LoadModule

This directive instructs the server to use the mod_ssl module.
The path is relative to the ServerRoot directive specified in the
httpd.conf configuration file. Without loading the module, our
encryption will not work.

98


5.4.3.2 Listen

This directive instructs the server to listen for incoming connec-
tions on port 443. This is the accepted convention for secure Web
communications (https). It is critical that this port be different
from the port used by the regular server. See the image above.

5.4.3.3 VirtualHost

Here, we define our secure Web server. Using the VirtualHost
block is the most elegant way of doing it. This allows you to create
additional blocks and serve additional secure sites to your clients,
allowing you an extra degree of flexibility and security.

99


Like we did before, we need to setup the DocumentRoot, the
ServerName and other directives. Let us review the most im-
portant elements:

<VirtualHost *:443>

This tells our server to listen on all interfaces for incoming connec-
tions on port 443. You may consider narrowing down the range to
speciﬁc IP addresses. Nevertheless, it is important to remember
that you can only use IP addresses! The secure Web server does
not permit named-based connections in its VirtualHost block. This
is because the SSL handshake occurs before the HTTP request can
identify the named-based virtual host.

100


Use only IP-based VirtualHost directives in the ssl.conf configuration
file! Name-based virtual hosts will fail.

DocumentRoot "/var/www/html"

This directive specifies the directory where all your web pages
should be stored. It is recommended that you use a different root
for non-secure and secure pages. However, in our example, we will
use the default selection. Just remember that this is NOT the
optimal setting.

ServerName www.ourserver.com:443

This entry defines the server name. If you do not use the hosts
file or DNS server for name resolution, you will have to specify an
IP address. We have solved this limitation earlier, so we can use
the server name here. In a production setup, where your server is
used by clients on the Internet, you will have to use DNS for name
resolution. For study and testing and in small, private networks,
the hosts file is an adequate solution.

This covers the first part of our setup. Now we must create the
certificate.

5.4.4 Create SSL certificate
Like we said before, we will create a CA, create a server key and
then sign the key with our self-created CA. In a production setup,

101


this will not work. If you intend to run any semi-serious business,
you will have to use a reputable, world-acknowledged CA to sign
your certificates.

Please note that the comparison between our setup and the real
scenario can be slightly confusing. If you get lost, there’s a ta-
ble summary (5.4.7) at the end of this section, emphasizing the
important differences between the two setups.

5.4.4.1 Create Certificate Authority (CA)

The first step is to create an encryption key, which we will use to
sign our CA. Please note that you should use a meaningful name
for the key. The best way to avoid confusion is to use the letters
ca in the name of the CA key. Likewise, use the word server when
creating the server key.

I have chosen the name myca.key, so that we do not confuse this
self-generated key (and the CA) with real keys.

102


Let us review the command:

openssl genrsa -des3 -out myca.key 4096

This OpenSSL command line tool will generate an RSA key, using
the Triple-DES cypher. The -out flag signifies the output name.
The number at the end of the command tells us how long the key
will be; generally, the longer the better. A 4096-bit encryption is
quite sufficient.

Please refer to openssl man page for more details.

After the key is created, you will be asked to use a password. This
means you won’t be able to use this key without providing the

103


password. While in theory, this is an interesting security measure,
it offers little actual benefit. We’ll discuss this soon.

Now that we have the key, we will create a CA.

openssl req -new -x509 -days 365 →
-key myca.key -out myca.crt

What do we have here? Well, basically, we are creating a certifi-
cate, using the key we have created earlier. Let us go over the
details:

104


req -new -x509

This part of the command tells us we want to issue a new X.509
Certificate Signing Request (CSR), where X.509 is an international
standard for public key and privilege management infrastructures.
In simple words, we want to create a certificate that will identify
our CA.

-days 365

This tells us how long the certificate will be valid. Security aspects
of this parameter are examined in greater depth in the Security
chapter (7.12).

-key myca.key -out myca.crt

We will use the key we have created earlier to sign the certificate
for the CA.

The command will invoke a guided text-interface wizard. We will
have to provide the password for the certificate key before we can
continue. After that, we will have to fill out an interactive form,
including the basic credentials that will identify us as the CA.

105


Please note that you should be careful when entering the Common
Name. You should use meaningful entries that will allow you to
easily distinguish your records, especially if you have several CAs.
Most people will never have to bother with this setting, but should
a need arise, here’s a pair of simple rules that you should adhere
to when creating CAs:

• For each CA, use the name of the site it will certify; in our case, ours-
erver.com (or www.ourserver.com).

• Append the letters CA to the end of the Common Name, so you will
know this is the CA entry.

In a real life situation, your credentials would be replaced with
those of an existing, reputable CA.

106


5.4.4.2 Create server key

We now have a certificate. It’s time to create the server key. The
principle is similar to what we’ve done before. The one thing
you should remember is that the server key should be named
server.key, in order to conform with Apache conventions.

After the encryption key is created, we will be asked to provide a
password to make the use of our key impossible without knowing
it. While this method is somewhat effective, it is not considered
a serious security measure. In fact, you are advised not to use it,
since the benefits do not outweigh the shortcomings.

Since you must provide the password any time the server is restarted
or reloaded, this means the secure server will not be able to start

107


after unattended reboot and will require a presence of an adminis-
trator to activate. This is cumbersome and can even be impracti-
cal. On the other hand, should your system be compromised, the
password will most likely present little challenge to the attacker.
Furthermore, compromised systems cannot be trusted, whether
passwords or other security methods are used.

Nevertheless, we will demonstrate both methods, so you can learn
and use both, should a need arise. We will begin with the password-
protected key and then later, create another one, which uses no
password.

5.4.4.3 Create Certiﬁcate Signing Request (CSR)

Now, we must “ask” our CA to sign our certiﬁcate. In a real
life situation, you would receive the server.csr from an existing,
established CA. Or you might even receive the signed key, with the
information you have provided in an application form, for instance.

Again, we must provide a password before we can continue. Then
again, we must go through an interactive form, providing details for
our site. In a real life situation, a real CA would ask you for these
details, whether via email, phone, an application form etc. For
more details, please refer to Submission of CSR to CA sub-section
below (5.5.2).

108


Since our CA and our website are one and the same, the form
will diﬀer little from what we have done when creating the CA.
This can be confusing. Therefore, you should remember that the
Common Name for your CA should include the letters CA (or
similar), to distinguish it from the server record.

Lastly, you can provide an additional password for the server key,
to make misuse more diﬃcult.

109


5.4.4.4 Sign Certificate Signing Request (CSR) with Certificate
Authority (CA)

What remains to be done is to sign the CSR with the CA we
have created. Once we do that, our certificate will be valid for the
coming year. After that, we will have to renew it.

110


You should be familiar with the syntax by now:

-CA myca.crt

This option instructs openssl to use our CA certiﬁcate.

-CAkey myca.key

This option instructs openssl to sign the certiﬁcate with the CA
key.

-set_serial 01

111


The set_serial option is used to create a serial number when
outputting a self-signed certificate. This allows you to track the
changes done to the certificate.

This covers the creation and signing of the SSL certificates.

5.4.4.5 Verify certificates

Let’s examine the certificates we have just created. This can help
you see if there are any problems with your files.

openssl rsa -noout -text -in myca.key

112


openssl x509 -noout -text -in myca.crt

113


openssl rsa -noout -text -in server.key

114


openssl x509 -noout -text -in server.crt

Everything looks good. Now, we can finish editing the ssl.conf file.

5.4.5 Edit ssl.conf configuration file - part 2
We now need to specify the location of our certificates and the
keys in the ssl.conf file so the server can find and use them. We
will comment out the sample entries in the file and use our own.
Necessarily, we will have to copy the relevant files to their right
location.

5.4.5.1 Server Certificate

This directive specifies the location of the server certificate (server.crt).
On CentOS 5, the default location is /etc/pki/tls/certs. We will

115


use the same directory. Your choice may vary. The important
thing to remember is to make the files unavailable to anyone but
root.

5.4.5.2 Server Private Key

This directive points to the location of the server key (server.key).
Again, your choice should reflect your needs. See image above.

5.4.5.3 Certificate Authority

This directive specifies the location of the CA certificate.

116


Now, let us copy the ﬁles to their relevant locations:

cp server.key /etc/pki/tls/private/server.key
cp server.crt /etc/pki/tls/certs/server.crt
cp myca.crt /etc/pki/tls/certs/myca.crt

We are ready. Let’s test our setup.

5.4.6 Test setup
After saving the ssl.conf ﬁle, we need to restart our server.


117


As you can see, we must provide a password before we can continue.
Now, we will try to access our server by typing
https://www.ourserver.com in the address line of a web browser.
You will most likely receive a warning message.

118


Let us examine the certiﬁcate before we accept it.

119


Indeed, everything looks fine. On the Web, though, very few peo-
ple would be convinced by this certificate. But in our setup, it
serves well. After accepting the certificate (either permanently or
temporarily for this session only), you will hit yet another warn-
ing.

120


This time, there’s a mismatch between the domain name
(www.ourserver.com) and the certiﬁcate (ourserver.com). This
should not be an issue if you are using the DNS server, but we will
discuss this separately in the next Part. For now, we will accept
the certiﬁcate.

After that, we should reach our site safely. Our setup works.

121


5.4.7 Mini-summary
Setting up the secure Web server might seem a little confusing.
Therefore, here’s a mini summary that should clarify the setup
process.

5.4.7.1 Names

This is a short list of file names used in this section:

File name Description
myca.key CA key
myca.crt CA certificate
server.key server key
server.csr server CSR
server.crt server certificate, signed by CA

122


5.4.7.2 Commands

Below, you can find a summarized list of commands you will need
to run to create your certificate. Please note that the names I
have used are generic and might not suit your needs. However, it
is important that you use the name server.key for the server key
file, to conform with Apache standards.

Command Description
openssl genrsa -des3 -out myca.key 4096 Create CA
key
openssl req -new -x509 -days 365 -key Create CA
myca.key -out myca.crt certificate
openssl genrsa -des3 -out server.key 4096 Create server
key
openssl req -new -key server.key -out Create CSR
server.csr
openssl x509 -req -days 365 -in server.csr Sign CSR
-CA myca.crt -CAkey myca.key -set_serial
01 -out server.crt

123


5.4.7.3 Difference between self-signed and CA-signed certificates

This will help you better understand the differences between our
exercise and a real, production setup.

Step Self-signed CA Real CA
Create CA Yes No need
key
Create CA Yes No need
Create CSR Yes As instructed by CA
Create server Yes Maybe:
key 1. CA creates key,
signs it and sends to
customer
2. CA creates CSR
only and sends to
customer, who then
creates server key and
signs with CA
Sign server Yes Maybe:
key 1. CA sends signed
key to customer
2. CA sends CSR;
customer signs the
key by himself/herself

124


5.4.7.4 Verification

You will have to run these commands to check your certificates and
keys:

openssl rsa -noout -text -in server.key
openssl x509 -noout -text -in server.crt
openssl rsa -noout -text -in myca.key
openssl x509 -noout -text -in myca.crt

5.4.7.5 File names and locations

These are the locations of relevant files:

Full path and name Description
/etc/httpd/conf.d/ssl.conf Main configuration file
/etc/httpd/modules Location of all modules
/etc/httpd/modules/mod_ssl.so Location of mod_ssl
module
/etc/pki/tls/certs Location of server and CA
certificates
/etc/pki/tls/private Location of server keys

125


5.5 Extras
Now that we have our secure Web server running, let us review a
number of other options.

5.5.1 Do not use password-protected server keys
As said before, the password protection for server keys is rather
impractical, without signiﬁcantly contributing to the server secu-
rity. Therefore, you are encouraged not to use them. Of course,
you must make sure that your server security is ensured by other
means.

5.5.1.1 Create server key without password

We will “ﬁlter” the old key (server.key) into a new one (server.key.nopass),
which will not include a password. Then, we will swap between the
old and the new one. You are advised to keep a backup copy of
the original key, just in case.

126


Next time we restart the server, we won’t be prompted for a pass-
word. All other settings remain unchanged.

127


5.5.2 Submission of CSR to CA
It is diﬃcult to say what method the CA will use to certify your
site. They might ask you for your credentials over a form or
through an online form. Or they might ask you to submit a CSR,
which they will sign, returning the key to you. Here, we will review
a typical option. Again, this may not be the case you’ll encounter.

5.5.2.1 Create CSR

Assuming the CA wants all records in the purely digital form, you
will have to create a CSR. We have already done that. Optionally,
they might ask you to convert the ﬁle into a Privacy Enhanced
Mail (PEM) format. Please note that this format is not widely
used and this will probably not be necessary.

128


openssl x509 -inform crt -in server.crt -out server.pem

Please note that you should consult your CA for detailed instruc-
tions regarding the conversions, if at all required.

5.5.2.2 Send CSR to CA

You will now have to submit the CSR file (or PEM) to the CA.
Once the CA processes your application, you will receive the cer-
tificate back. The file will most likely be sent in the PEM format,
so you will have to convert it back to CSR format. It will also
most likely bear a different name from what you are used to, so
you should rename it to server.crt, to conform with Apache con-
ventions.

5.5.2.3 Verify certificate

Now, you should verify the certificate, against the relevant CA’s
file. You should receive this file from your CA. Alternatively, if
your distro includes a list of CAs, you might try that one. On
CentOS 5, a bundle containing a list of known CAs is located under
/etc/pki/tls/certs.

openssl verify -CAfile ca-bundle.crt →
-purpose sslserver server.crt

Next, you should verify that the certificate corresponds to the pri-
vate key. Please make sure the names match.

openssl x509 -noout -modulus -in server.pem | openssl md5
openssl rsa -noout -modulus -in server.crt | openssl md5

129


Once you have completed the above steps, you will have to edit the
ssl.conf file, restart the server and test your setup. We’re back on
familiar grounds.

For more details regarding the different commands and verification,
please consult:

• verify man page

• OpenSSL: Documents, req

5.6 General considerations
Here’s a number of settings that you should remember.

5.6.1 Use secure server only
If you want your server to serve only secure pages, simply comment
the Listen 80 directive from the httpd.conf file. Sites dedicated to
clients privacy and security should not run any other, non-secure
content.

5.6.2 Use only IP-based virtual hosts
It is impossible to use named-based virtual hosts with the secure
Web server. This is because the SSL handshake occurs before the
HTTP request can identify the named-based virtual host. Using
names will result in errors. You may only use IP addresses in the
VirtualHost directives inside the ssl.conf configuration file.

130


5.6.3 Use server.key as ﬁle name for the server key
It is not strictly necessary, but it is good practice and in line with
Apache conventions to use the server.key name for the server key
ﬁle. If for some reason you require paranoid security, then you
might change this name to something less obvious.

131

Chapter 6

Other configurations

6.1 Firewall rules
Here’s the most basic pair of rules to allow HTTP traffic (including
both secure and non-secure):

iptables -A INPUT -p tcp –dport 80 -j ACCEPT
iptables -A INPUT -p tdp –dport 443 -j ACCEPT

However, you can restrict the traffic even more. For example, you
can allow incoming connections only to a certain interface:

iptables -A INPUT -p tcp –dport 80 -i eth0 -j ACCEPT
iptables -A INPUT -p tdp –dport 443 -i eht0 -j ACCEPT

132


Finally, you can restrict the traffic to a specific subnet, allowing
only certain machines to connect:

iptables -A INPUT -p tcp –dport 80 -i eth0 →
-s 192.168.1.0/24 -j ACCEPT
iptables -A INPUT -p tcp –dport 443 -i eth0 →
-s 192.168.1.0/24 -j ACCEPT

You can further sharpen the rules by specifying source ports and
packet states. We’ll see this in the next section.

6.1.1 Advanced firewall rules
Sometimes, your Web server might not be an external client; it
will connect to the Internet through a dedicated, firewalled gate-
way. This brings about several issues, which include the Network
Address Translation (NAT) and port forwarding.

We have discussed the forwarding and masquerading in Part ?: Dy-
namic Host Configuration Protocol (DHCP) server - ISC DHCPD
(). Now, we will look at a more complex setup.

Let’s assume that your Web server is a local machine, with a local
IP address. It serves both internal and external clients. The inter-
nal setup is rather simple. We need to make sure external clients
can connect, too.

133


We will need to allow traffic destined to ports 80 and 443 of our
external IP address (let’s assume 1.1.1.1) to be forwarded to ports
80 and 443 of our Web server, which resides on a local address
(192.168.1.128). Furthermore, interfaces are eth0 for the Internet
and eth1 for the intranet.

Our setup will include several steps:

• We will have to “forward” our web ports, so that clients behind the
gateway will be able to accept incoming communications.

• We will have to setup some sort of masquerading, only this time we
will use methods slightly different from those we have adopted when
setting up the DHCP server.

6.1.1.1 Port forwarding

Our gateway is configured to forward communications between in-
ternal and external hosts, with new, established and related con-
nections permitted outbound and established and related connec-
tions permitted inbound. This is a great setup for a server-less
network, but it won’t do for Apache.

Here are the original rules, which we have setup for the DHCP
server acting as a gateway:

iptables -A FORWARD -t filter -i eth1 -m state →
–state ESTABLISHED,RELATED -j ACCEPT

iptables -A FORWARD -t filter -o eth1 -m state →
–state NEW,ESTABLISHED,RELATED -j ACCEPT

134


Now, we need these rules to make it work:

iptables -I FORWARD -p tcp -i eth1 -o eth0 →
-d 192.168.1.128 –dport 80 →
-m state –state NEW -j ACCEPT

-d 192.168.1.128 –dport 443 →

If you want to tighten the rules some more, you can also specify
the source ports:

-d 192.168.1.128 –dport 80 -sport 1024:65535 →

We have placed the rules on the top of the chain, so they would be
processed before the existing rules, which only allow new outbound
connections. Basically, these rules are sufficient if your gateway is
servicing a number of local networks, all of which can fully resolve
one another’s IP addresses. They are not good enough for the
Internet, though.

6.1.1.2 Destination NAT

Using masquerading the way we did when we configured the DHCP
server might not be good enough for us, because it will point all
traffic to the default network interface. We have to use a more
sophisticated method, which is the DNAT. Now, we need to allow

135


new incoming connections to our Web server to properly resolve
to the right client, on the right ports.

iptables -t nat -A PREROUTING -p tcp -i eth1 -d 1.1.1.1 →
–dport 80 –to-destination 192.168.1.128:80 -j DNAT

iptables -t nat -A PREROUTING -p tcp -i eth1 -d 1.1.1.1 →
–dport 443 –to-destination 192.168.1.128:443 -j DNAT

Of course, you can use non-default ports on the Web server, like
8080 or anything alike, which makes the idea of port forwarding
even more meaningful.

To reiterate, IP masquerading is good enough for “normal” brows-
ing, but servers behind the firewall also require port forwarding.
Casual peer-to-peer (P2P) home users behind routers often have
to do this to make their programs work.

6.1.1.3 Static NAT

If you have more than one publicly visible IP address, you won’t
be able to use IP masquerading. This is because masquerading
forces all traffic to the default network interface on the firewalled
gateway, resulting in a single usable external IP address.

However, it is quite likely that you will want to run your servers on
separate hosts, with different both internal and external addresses,
both to shape your traffic in a more orderly fashion and reduce the
workload on specific hosts. To this end, you will have to use SNAT
rather than IP forwarding.

136


The basic principle remains the same, except that you use sepa-
rate external IP addresses for individual hosts, groups of hosts or
the entire local network, as you see ﬁt. In our example, we will
demonstrate SNAT by creating a private rule for the Web server
and a general rule for all other clients.

Let’s assume the Web server will use a public IP address of 1.1.1.1,
while all other clients will use 1.1.1.2.

137


Web server rules:

These two rules are required to allow DNAT and SNAT for the
client running the Web server. Please note that these two rules
do not specify what kind of servers are running on the particular
client. This grants you extra ﬂexibility, if you need to run more
than one server on a particular machine.

iptables -t nat -A PREROUTING -d 1.1.1.1 -i eth1 →
–to-destination 192.168.1.128 -j DNAT

iptables -t nat -A POSTROUTING -s 192.168.1.128 -o eth1 →
–to-source 1.1.1.1 -j SNAT

As said, the forwarding rules from before remain valid, both the
speciﬁc rules for the Web server, which permit new inbound con-
nections, and the general rules, which permit only new outbound
connections.

138


General rules:

This rule applies to all local network clients, trying to communicate
with the external network.
iptables -t nat -A POSTROUTING -s! 192.168.1.128 -o eth1 →
–to-source 1.1.1.2 -j SNAT

6.2 Enable Web server on startup
You will most likely want your Apache server to run on startup.
The simplest way to enable this is to use the chkconﬁg utility.

chkconfig –levels 5 httpd on

139

Chapter 7

Security

Web server security is one of the most important things in your
setup. If your server becomes compromised, you run the risk of
serving malicious, fraudulent or simply tasteless pages to hundreds
and thousands of your visitors. Furthermore, you risk exposing the
privacy of your clients and users. Forum names and passwords,
email addresses, sensitive records, credit card numbers, and other
information could all be leaked out, creating an identity theft night-
mare.

It is paramount that you keep your Web server in tiptop shape at
all times. This requires lots of work, attention and responsibility
and is not a trivial task. Running a good server takes time and
patience.

It is also important that you be constantly aware of what goes
about on your server. If you have several users uploading material
to their individual directories, you are advised to make sure that
they do not post potentially dangerous content. Most Linux users
are oblivious to the web threats, but a large percentage of Windows

140


users have a hard time with sites loaded with malicious payload.
As the server owner and administrator, it is your responsibility to
make sure that your visitors are not at risk.

Let us review some of the most crucial settings that you should
pay attention to make sure both your server and your clients are
secure.

7.1 Updates
Keep your server up to date at all times. Make sure you patch
new vulnerabilities instantly. You are advised to subscribe to the
Apache HTTP Server Mailing Lists for information about new
bugs, updates, features, and more.

7.2 Hide your server version
This is a “security through obscurity” measure. Nevertheless, it
does not hurt to use it. At the very least, this will annoy and
delay a potential attacker, by making his attempts to harvest server
information more difficult.

To remove server information, you will need to use these two di-
rectives:

ServerSignature Off
ServerTokens Prod

The first directive, ServerSignature, will remove the server version
information from the pages generated by the server, like error pages

141


(403 Forbidden, 404 Not found), directory listings and others. The
second directive, ServerTokens, will change the server’s HTTP
Response Header. By default, with the directive set to OS, the
header will disclose both the version and the operating system.
Set to Prod, the header will merely report Apache.

Here’s an example without these directives.

142


And here’s with the directives in place. The version is disguised.

7.3 Logs
You should check your logs at least daily. This may be tedious and
boring, but it is vital that you discover any potential breaches as
quickly as possible. Keep an eye on things and look for suspicious
directories and files.

7.4 Permissions
Badly implemented permissions can ruin your entire security. It is
critical that you make sure the executables, configuration files, logs,
access files and private keys are located outside the public HTML
directories and writable only by root. The web pages should be

143


readable and possibly executable by your visitors but only writable
by their respective owners.

Let us review the necessary permissions:

Location Permissions
/usr/sbin/httpd F: 511
/etc/httpd D: 751
/etc/httpd/conf D: 751, F: 644 / 600
/etc/httpd/conf.d D: 751, F: 644 / 600
/etc/httpd/logs symbolic link, 755 / 711
/etc/httpd/modules symbolic link, 755 / 711
/etc/httpd/run symbolic link, 755 / 711
/usr/lib/httpd/modules D: 751, F: 644 / 600
/var/log/httpd D: 751, F: 644 / 600
/var/run D: 751, F: 644 / 600
/var/www/html D: 755, F: 755

If you are really paranoid, then you should ONLY allow root access
to the binaries and configuration files. It really depends on your
setup and needs. Last but not the least, let’s not forget that system
files MUST be owned by root.

144


7.5 Access to root (/)
You must not allow anyone to access the root directory. Therefore,
you should implement a default deny policy for the root and all sub-
directories and then partially allow access to specific locations, like
the public HTML directories of your users.

<Directory />
Order Deny, Allow
Deny from all
</Directory>

7.6 AllowOverride
This directive specifies if options used in the .htaccess files can
conflict (and thus override) the settings configured for the particu-
lar directory. In general, you should set this directive to none and
only permit specific tasks to a small number of trusted users. If
you lease your server to numerous clients who must have some sort
of protection for their content, then you can allow them to use the
.htaccess file for authentication, as we have shown before (4).

AllowOverride none

145


7.7 Disable public access to .ht files
You must not allow any public user to be able to load the .htaccess
file through the browser window. Furthermore, the .htpasswd file,
which contains the user names and passwords, must also be pro-
tected from public access. Again, we have discussed this before,
but it doesn’t hurt to repeat it.

<Files ~ “^.ht”>
Order allow, deny
Deny from all
</Files>

7.8 Dynamic content
You need to be very careful with dynamic pages and scripts, since
they allow server to perform a variety of operations that static
HTML files cannot do. You should think twice before you allow any
user to run scripts from his/her own public directory. In general,
scripts are only allowed in special directories, as defined by the
ScriptAlias directive.

7.8.1 Disable CGI
You should disable scripts in user directories. If you really must
permit them, use the Options +ExecCGI directive.

Options -ExecCGI

146


7.8.2 Disable Server Side Includes (SSI)
SSI is a simple scripting language that allows web servers to display
variables or execute other programs from within web pages. This
introduces a significant load on the server and poses a security
risk. You should not allow SSI to function on your server unless
absolutely necessary.

Options -Includes

For more about SSI, please refer to Server Side Includes on Wikipedia.

7.9 Disable unnecessary modules
By default, Apache is very permissive when it comes to the func-
tionality it offers. You should comment out any modules you do
not need or wish to use from the /etc/httpd/conf/httpd.conf file.

7.10 Use ModSecurity (mod_security) mod-
ule
ModSecurity is a powerful Web Application Filter (WAF), which
allows you to greatly enhance the security of your server by to de-
tecting and preventing attacks before they reach web applications.
ModSecurity can perform a variety of tasks, including detection of
HTTP protocols violation, detection against common web attacks,
detection of bots and crawlers, detection of Trojan horses, filter-
ing based on existing rulesets, policies or regular expressions, and
more. The application relies on generic rules to detect and prevent
exploits and does not rely on blacklists or signatures.

147


Furthermore, it will not throttle the traffic throughput. And best
of all, ModSecurity is very easy add to an existing and running
Apache server.

Setting up and configuring ModSecurity is outside the scope of
this Part. However, it will be reviewed separately, in the context
of Network Security. If you are interested, you can read Part ?,
Chapter ?: ModSecurity ().

Meanwhile, for more details, you should refer to ModSecurity:
Open Source Web Application Firewall.

7.11 Chroot Jail
Like with many other services, it is possible to “sandbox” httpd to
run in a virtual prison. The service will think its files are located in
the default directories. In reality, they will reside inside a Chroot
Jail, which will mimic the layout of the real directories, except
that many of the files normally found under the real root won’t be
there, preventing possible privilege escalation risks.

This is an extremely important security feature and should be used
whenever possible. For more details about Chroot Jail, please refer
to Part ?: Chroot Jail, Chapter ?: Apache in Chroot Jail () for
details how to create this setup.

For more information, you are also welcome to read: Apache in a
chroot jail.

148


7.12 Secure web server only
The secure Web servers will offer their certificates to any client that
asks for them. This means that you do not care about who your
clients are. However, if your setup also requires that only a limited
number of clients be allowed to access the secure content, you
might consider using any one or several of the following methods:
Kerberos, firewall rules, TCP wrappers, allow & deny directives,
.htaccess files, client certificates, and more.

7.12.1 Different DocumentRoot
You are advised to use a separate directory for the secure pages.
This allows you to fully separate normal content from secure, priv-
ileged data and might even mitigate potential exposure in case of
an attack.

7.12.2 Permissions
The server.crt and the server.key file must only be readable by
root. You should even disallow the root user from making any
changes to the files. Set the permissions for these two files to 0400.

The permissions for the ssl.conf should be in line with your policy,
which should be either 640 or 600.

7.12.3 Duration of certificates
Normally, certificates are issued for a year. However, if you are
running an ultra-paranoid setup, you may want to make the cer-
tificates expire after only a few weeks. This means that potential
attackers will always have only a limited access to data for a short

149


period of time, in case they succeed in decrypting your ciphers.
You will force them to work all over again, trying to decode your
data. Needless to say, you will most likely use self-signed certifi-
cates for this type of work.

7.13 Word of caution
This Part of the book cannot possibly encompass all the security
aspects of running Apache. Do not presume you know everything
about running a Web server just by reading a few pages here and
there - I sure do not. Keep in mind that you are responsible for ev-
ery bit of traffic flowing to and from your server - and then branch
from there. Updates and logs are your best friends. Then, there’s
the matter of ethics. Your websites could be servings hundreds,
thousands or even millions of people every day. The choices you
make regarding your content and security policies may affect all of
them.

Needless to say, nothing can replace years of hard work and expe-
rience.

150

Chapter 8

Additional resources

Here, you will ﬁnd a number of additional links to useful resources
that can help you better learn and understand the Apache Web
server. You are warmly encouraged to visit them.

• The Apache Software Foundation

• The Apache HTTP Server Project

• The Apache Interface to OpenSSL

• Security Tips - Apache HTTP Server

151

Chapter 9

Exercises

This section is meant to help you estimate your knowledge of the
Apache Web server, now that you have read the Part and worked
through every step (you have, haven’t you?). If you can successfully
answer all of the questions below with ease, this probably means
you possess a decent knowledge of the Apache Web server and
you can most likely create a setup of your own without too much
diﬃculty.

Please note that this short questionnaire is by no means an exhaus-
tive or any sort of format test to Apache expertise. It should be a
good starting point for broadening your knowledge and experience.

When trying to solve the questions, please do not look into the
answers right away. Try to think and ﬁgure out the problem on
your own. The questions might also help you identify the stronger
and weaker points in your freshly earned Apache skills.

Furthermore, you are encouraged to simulate the problems by pow-
ering up Linux and practicing for real.

152


9.1 Questions
9.1.1 Secure Web server & VirtualHost
You want to configure a secure Web server that you will use for
online transactions with your customers. You have completed all
the steps successfully, but your server does not work, for some
reason. Looking at the VirtualHost block in the ssl.conf file, you
are unable to find the source of the problem. Can you tell what’s
wrong?

Please note that there is an error only within the displayed di-
rectives in the screenshot. The VirtualHost container is properly
closed and all directives have the correct syntax.

153


9.1.2 Directory, Files and Locations
You have used the AllowOverride directive inside a Location block
to setup the behavior of the .htaccess file. However, your server
ignores the change. What might be the reason?

<Location /dir/>
Order allow, deny
Allow from all
Deny from plush-mush.info
AllowOverride AuthConfig
</Location>

1. AllowOverride must be placed first inside any block.

2. AllowOverride can only be used inside Directory tags.

3. You have not restarted the server.

4. You need to specify the location of the .htaccess file in the opening
Location tag.

9.1.3 Server functionality, 1
You want to use CGI scripts on your pages, but the server simply
refuses to comply. You are absolutely convinced your syntax is
correct. What could be the likely reason for your woes?

154


Something is horribly wrong with your httpd.conf ﬁle. Nothing
works. What’s the matter?

155


9.1.5 .htaccess
What security precautions must you take to make sure authenti-
cation via .htaccess is not compromised easily?

1. Place the .htpasswd in a restricted directory inaccessible from the web.

2. Place the .htaccess ﬁle in a restricted directory and chmod it to 755.

3. Use a Files or FilesMatch directive to restrict access to .ht ﬁles.

4. Use a reasonably strong password for user authentication.

5. Change the Listen 80 directive to a random number.

156


9.2 Answers
9.2.1 Secure Web server & VirtualHost
It is impossible to use named-based virtual hosts with the secure
Web server. This is because the SSL handshake occurs before the
HTTP request can identify the named-based virtual host. Using
names will result in errors. You may only use IP addresses in the
VirtualHost directives inside the ssl.conf configuration file.

9.2.2 Directory, Files and Locations
The right answer is 2. If you have read the Apache documentation,
as suggested in the links provided, you would have realized this
critical detail. The AllowOverride directive can only used inside
Directory tags. However, the right answer might also be derived by
running a few simple tests. Answer 3 is the trickiest, because it implies
one of the classic mistakes. However, running a simple test will verify this
is not the case. Answer 1 is misleading, especially since the Order directive
was used - for which, the precedence does matter. Again, a quick test will
prove this false. The last answer is pure nonsense.

The CGI functionality is provided by the mod_actions module.
You need to check that the httpd.conf file contains a LoadModule
directive for this module.

157


Furthermore, you need to check that the module can indeed be
found under: /usr/lib/httpd/modules.

You have placed a trailing slash at the end of the ServerRoot di-
rective. This will cause all your symbolic links to fail. In the ﬁle
comments that have been removed (a BAD idea!), there was a no-
tice warning against placing the trailing slash at the end of the
directive.

You should never add a slash at the end of the directory path -
and ever delete comments.

158


9.2.5 .htaccess
The right answers are 1, 3 and 4. The .htaccess file must be
placed in public directories, thus placing them in non-web direc-
tories would invalidate their purpose. This makes answer no. 2
wrong. Furthermore, the executable bit in the permissions is com-
pletely unnecessary for the functionality of the file.

The connection to the Web server (and the relevant port) is estab-
lished before any authentication takes place. Moreover, the choice
of the port number has no bearing whatsoever on the functionality
of the .htaccess files. Thus, answer no. 5 is wrong.

159

Apache Web server Complete Guide

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to Apache Web server Complete Guide (20)

More from webhostingguy (20)

Apache Web server Complete Guide