Installing mod_security
Modsecurity is available in the Debian/Ubuntu repository:
apt-get install libapache2-modsecurity
Verify if the mod_security module was loaded.
apachectl -M | grep --color security
You should see a module named security2_module (shared)
which indicates that the module was loaded.
Modsecurity’s installation includes a recommended configuration file which has to be renamed:
mv /etc/modsecurity/modsecurity.conf{-recommended,}
Reload Apache
service apache2 reload
You’ll find a new log file for mod_security in the Apache log directory:
root@droplet:~# ls -l /var/log/apache2/modsec_audit.log
-rw-r----- 1 root root 0 Oct 19 08:08 /var/log/apache2/modsec_audit.log
Configuring mod_security
Out of the box, modsecurity doesn’t do anything as it needs rules to work. The default configuration file is set to DetectionOnly which logs requests according to rule matches and doesn’t block anything. This can be changed by editing the modsecurity.conf
file:
nano /etc/modsecurity/modsecurity.conf
Find this line
SecRuleEngine DetectionOnly
and change it to:
SecRuleEngine On
If you’re trying this out on a production server, change this directive only after testing all your rules.
Another directive to modify is SecResponseBodyAccess
. This configures whether response bodies are buffered (i.e. read by modsecurity). This is only neccessary if data leakage detection and protection is required. Therefore, leaving it On will use up droplet resources and also increase the logfile size.
Find this
SecResponseBodyAccess On
and change it to:
SecResponseBodyAccess Off
Now we’ll limit the maximum data that can be posted to your web application. Two directives configure these:
SecRequestBodyLimit
SecRequestBodyNoFilesLimit
The SecRequestBodyLimit
directive specifies the maximum POST data size. If anything larger is sent by a client the server will respond with a 413 Request Entity Too Large error. If your web application doesn’t have any file uploads this value can be greatly reduced.
The value mentioned in the configuration file is
SecRequestBodyLimit 13107200
which is 12.5MB.
Similar to this is the SecRequestBodyNoFilesLimit
directive. The only difference is that this directive limits the size of POST data minus file uploads– this value should be “as low as practical.”
The value in the configuration file is
SecRequestBodyNoFilesLimit 131072
which is 128KB.
Along the lines of these directives is another one which affects server performance: SecRequestBodyInMemoryLimit
. This directive is pretty much self-explanatory; it specifies how much of “request body” data (POSTed data) should be kept in the memory (RAM), anything more will be placed in the hard disk (just like swapping). Since droplets use SSDs, this is not much of an issue; however, this can be set a decent value if you have RAM to spare.
SecRequestBodyInMemoryLimit 131072
This is the value (128KB) specified in the configuration file.
Testing SQL Injection
Before going ahead with configuring rules, we will create a PHP script which is vulnerable to SQL injection and try it out. Please note that this is just a basic PHP login script with no session handling. Be sure to change the MySQL password in the script below so that it will connect to the database:
/var/www/login.php
<html>
<body>
<?php
if(isset($_POST['login']))
{
$username = $_POST['username'];
$password = $_POST['password'];
$con = mysqli_connect('localhost','root','password','sample');
$result = mysqli_query($con, "SELECT * FROM `users` WHERE username='$username' AND password='$password'");
if(mysqli_num_rows($result) == 0)
echo 'Invalid username or password';
else
echo '<h1>Logged in</h1><p>A Secret for you....</p>';
}
else
{
?>
<form action="" method="post">
Username: <input type="text" name="username"/><br />
Password: <input type="password" name="password"/><br />
<input type="submit" name="login" value="Login"/>
</form>
<?php
}
?>
</body>
</html>
This script will display a login form. Entering the right credentials will display a message “A Secret for you.”
We need credentials in the database. Create a MySQL database and a table, then insert usernames and passwords.
mysql -u root -p
This will take you to the mysql>
prompt
create database sample;
connect sample;
create table users(username VARCHAR(100),password VARCHAR(100));
insert into users values('jesin','pwd');
insert into users values('alice','secret');
quit;
Open your browser, navigate to http://yourwebsite.com/login.php
and enter the right pair of credentials.
Username: jesin
Password: pwd
You’ll see a message that indicates successful login. Now come back and enter a wrong pair of credentials– you’ll see the message Invalid username or password.
We can confirm that the script works right. The next job is to try our hand with SQL injection to bypass the login page. Enter the following for the usernamefield:
' or true --
Note that there should be a space after --
this injection won’t work without that space. Leave the password field empty and hit the login button.
Voila! The script shows the message meant for authenticated users.
Setting Up Rules
To make your life easier, there are a lot of rules which are already installed along with mod_security. These are called CRS (Core Rule Set) and are located in
root@droplet:~# ls -l /usr/share/modsecurity-crs/
total 40
drwxr-xr-x 2 root root 4096 Oct 20 09:45 activated_rules
drwxr-xr-x 2 root root 4096 Oct 20 09:45 base_rules
drwxr-xr-x 2 root root 4096 Oct 20 09:45 experimental_rules
drwxr-xr-x 2 root root 4096 Oct 20 09:45 lua
-rw-r--r-- 1 root root 13544 Jul 2 2012 modsecurity_crs_10_setup.conf
drwxr-xr-x 2 root root 4096 Oct 20 09:45 optional_rules
drwxr-xr-x 3 root root 4096 Oct 20 09:45 util
The documentation is available at
root@droplet1:~# ls -l /usr/share/doc/modsecurity-crs/
total 40
-rw-r--r-- 1 root root 469 Jul 2 2012 changelog.Debian.gz
-rw-r--r-- 1 root root 12387 Jun 18 2012 changelog.gz
-rw-r--r-- 1 root root 1297 Jul 2 2012 copyright
drwxr-xr-x 3 root root 4096 Oct 20 09:45 examples
-rw-r--r-- 1 root root 1138 Mar 16 2012 README.Debian
-rw-r--r-- 1 root root 6495 Mar 16 2012 README.gz
To load these rules, we need to tell Apache to look into these directories. Edit the mod-security.conf
file.
nano /etc/apache2/mods-enabled/mod-security.conf
Add the following directives inside <IfModule security2_module> </IfModule>
:
Include "/usr/share/modsecurity-crs/*.conf"
Include "/usr/share/modsecurity-crs/activated_rules/*.conf"
The activated_rules
directory is similar to Apache’s mods-enabled
directory. The rules are available in directories:
/usr/share/modsecurity-crs/base_rules
/usr/share/modsecurity-crs/optional_rules
/usr/share/modsecurity-crs/experimental_rules
Symlinks must be created inside the activated_rules
directory to activate these. Let us activate the SQL injection rules.
cd /usr/share/modsecurity-crs/activated_rules/
ln -s /usr/share/modsecurity-crs/base_rules/modsecurity_crs_41_sql_injection_attacks.conf .
Apache has to be reloaded for the rules to take effect.
service apache2 reload
Now open the login page we created earlier and try using the SQL injection query on the username field. If you had changed the SecRuleEngine
directive toOn, you’ll see a 403 Forbidden error. If it was left to the DetectionOnly option, the injection will be successful but the attempt would be logged in the modsec_audit.log
file.
Writing Your Own mod_security Rules
In this section, we’ll create a rule chain which blocks the request if certain “spammy” words are entered in a HTML form. First, we’ll create a PHP script which gets the input from a textbox and displays it back to the user.
/var/www/form.php
<html>
<body>
<?php
if(isset($_POST['data']))
echo $_POST['data'];
else
{
?>
<form method="post" action="">
Enter something here:<textarea name="data"></textarea>
<input type="submit"/>
</form>
<?php
}
?>
</body>
</html>
Custom rules can be added to any of the configuration files or placed in modsecurity directories. We’ll place our rules in a separate new file.
nano /etc/modsecurity/modsecurity_custom_rules.conf
Add the following to this file:
SecRule REQUEST_FILENAME "form.php" "id:'400001',chain,deny,log,msg:'Spam detected'"
SecRule REQUEST_METHOD "POST" chain
SecRule REQUEST_BODY "@rx (?i:(pills|insurance|rolex))"
Save the file and reload Apache. Open http://yourwebsite.com/form.php
in the browser and enter text containing any of these words: pills, insurance, rolex.
You’ll either see a 403 page and a log entry or only a log entry based on SecRuleEngine
setting. The syntax for SecRule is
SecRule VARIABLES OPERATOR [ACTIONS]
Here we used the chain action to match variables REQUEST_FILENAME withform.php, REQUEST_METHOD with POST and REQUEST_BODY with the regular expression (@rx) string (pills|insurance|rolex). The ?i: does a case insensitive match. On a successful match of all these three rules, the ACTIONis to deny and log with the msg “Spam detected.” The chain action simulates the logical AND to match all the three rules.
Excluding Hosts and Directories
Sometimes it makes sense to exclude a particular directory or a domain name if it is running an application like phpMyAdmin as modsecurity and will block SQL queries. It is also better to exclude admin backends of CMS applications like WordPress.
To disable modsecurity for a complete VirtualHost place the following
<IfModule security2_module>
SecRuleEngine Off
</IfModule>
inside the <VirtualHost>
section.
For a particular directory:
<Directory "/var/www/wp-admin">
<IfModule security2_module>
SecRuleEngine Off
</IfModule>
</Directory>
If you don’t want to completely disable modsecurity, use the SecRuleRemoveById
directive to remove a particular rule or rule chain by specifying its ID.
<LocationMatch "/wp-admin/update.php">
<IfModule security2_module>
SecRuleRemoveById 981173
</IfModule>
</LocationMatch>
Further Reading
Official modsecurity documentationhttps://github.com/SpiderLabs/ModSecurity/wiki/Reference-Manual