Documentation Home
Security in MySQL
Related Documentation Download this Excerpt
PDF (US Ltr) - 2.0Mb
PDF (A4) - 2.0Mb
HTML Download (TGZ) - 429.6Kb
HTML Download (Zip) - 436.8Kb

Security in MySQL  /  ...  /  Using MySQL Enterprise Firewall

6.7.3 Using MySQL Enterprise Firewall

Before using MySQL Enterprise Firewall, install it according to the instructions provided in Section 6.7.2, “Installing or Uninstalling MySQL Enterprise Firewall”.

This section describes how to configure MySQL Enterprise Firewall using SQL statements. Alternatively, MySQL Workbench 6.3.4 or higher provides a graphical interface for firewall control. See MySQL Enterprise Firewall Interface.

Enabling or Disabling the Firewall

To enable or disable the firewall, set the mysql_firewall_mode system variable. By default, this variable is enabled when the firewall is installed. To control the initial firewall state explicitly, you can set the variable at server startup. For example, to enable the firewall in an option file, use these lines:


After modifying my.cnf, restart the server to cause the new setting to take effect.

Alternatively, to set and persist the firewall setting at runtime:

SET PERSIST mysql_firewall_mode = OFF;
SET PERSIST mysql_firewall_mode = ON;

SET PERSIST sets the value for the running MySQL instance. It also saves the value, causing it to carry over to subsequent server restarts. To change a value for the running MySQL instance without having it carry over to subsequent restarts, use the GLOBAL keyword rather than PERSIST. See SET Syntax for Variable Assignment.

Assigning Firewall Privileges

With the firewall installed, grant the appropriate privileges to the MySQL account or accounts that are expected to administer it. The privileges depend on which firewall operations an account should be permitted to perform:

  • Grant the FIREWALL_ADMIN privilege to any account that should have full adminstrative firewall access. (Some user-defined procedures can be invoked by accounts that have FIREWALL_ADMIN or the deprecated SUPER privilege, as indicated in individual UDF descriptions.)

  • Grant the FIREWALL_USER privilege to any account that should have administrative access only for its own firewall rules.

  • Grant the EXECUTE privilege for the firewall stored procedures in the mysql system database. These may invoke UDFs, so stored procedure access also requires the privileges indicated earlier that are needed for those UDFs.


The FIREWALL_ADMIN and FIREWALL_USER privileges can be granted only while the firewall is installed because the firewall component defines those privileges.

Firewall Operational Concepts

The MySQL server permits clients to connect and receives from them SQL statements to be executed. The server passes to the firewall each incoming statement that does not immediately fail with a syntax error. Based on whether the firewall accepts the statement, the server executes it or returns an error to the client.

Firewall operation is based on a registry of profiles that enable statement execution protection to be applied. A profile has these attributes:

  • Rules that define which statements are acceptable to the profile. This set of rules forms the profile allowlist.

  • The current operational mode. The mode enables the profile to be used in different ways. For example: the profile can be placed in training mode to establish the allowlist; the allowlist can be used for restricting statement execution or intrusion detection; the profile can be disabled entirely.

  • The scope of applicability, indicating which client connections the profile applies to.

    The firewall supports account-based profiles such that each profile matches a particular client account (client user name and host name combination). For example, you can register one account profile for which the allowlist applies to connections originating from admin@localhost and another account profile for which the allowlist applies to connections originating from

For each client connection, the firewall determines which profile applies, and accepts only statements the profile allowlist permits. (If the client matches no profile, the firewall ignores it and accepts all statements.)

By default, the firewall accepts all statements and has no effect on which statements MySQL accounts can execute. To apply firewall protective capabilities, you must take explicit action:

  • Register one or more profiles with the firewall. (This is required because the firewall ignores clients that match no profile.)

  • Train the firewall to establish the allowlist for each profile; that is, the types of statements the profile permits clients to execute.

  • Tell the firewall to protect MySQL using each profile for which it has been trained; that is, to match incoming statements against the appropriate allowlist when clients connect.

Statement matching performed by the firewall does not use SQL statements as received from clients. Instead, the server converts incoming statements to normalized digest form and firewall operation uses these digests. The benefit of statement normalization is that it enables similar statements to be grouped and recognized using a single pattern. For example, these statements are distinct from each other:

SELECT first_name, last_name FROM customer WHERE customer_id = 1;
select first_name, last_name from customer where customer_id = 99;
SELECT first_name, last_name FROM customer WHERE customer_id = 143;

But all of them have the same normalized digest form:

SELECT `first_name` , `last_name` FROM `customer` WHERE `customer_id` = ?

By using normalization, the firewall can store digests in allowlists that each match many different statements received from clients. For more information about normalization and digests, see Performance Schema Statement Digests and Sampling.

Each profile registered with the firewall has its own operational mode, chosen from these values:

  • OFF: This mode disables the profile. The firewall considers it inactive and ignores it.

  • RECORDING: This is the firewall training mode. Incoming statements received from a client that matches the profile are considered acceptable for the profile and become part of its fingerprint. The firewall records the normalized digest form of each statement to learn the acceptable statement patterns for the profile. Each pattern is a rule, and the union of the rules is the profile allowlist.

  • PROTECTING: In this mode, the profile allows or prevents statement execution. The firewall matches incoming statements against the profile allowlist, accepting only only statements that match and rejecting those that do not. After training the profile in RECORDING mode, switch it to PROTECTING mode to harden MySQL against access by statements that deviate from the allowlist.

  • DETECTING: This mode detects but not does not block intrusions (statements that are suspicious because they match nothing in the profile allowlist). In DETECTING mode, the firewall writes suspicious statements to the error log but accepts them without denying access.

When a profile is assigned any of the preceding mode values, the firewall stores the mode in the profile. Firewall mode-setting operations also permit a mode value of RESET, but this value is not stored: setting a profile to RESET mode causes the firewall to delete all rules for the profile and set its mode to OFF.

Registering Firewall Account Profiles

MySQL Enterprise Firewall enables profiles to be registered that correspond to individual accounts.

MySQL authenticates each client session for a specific user name and host name combination. This combination is the session account. The firewall matches the session account against registered account profiles to determine which profile applies to handling incoming statements from the session:

  • The firewall ignores inactive profiles (profiles with a mode of OFF).

  • The session account matches the active account profile having the same user and host, if there is one. There is at most one such account profile.

In other words, at most one active account profile is applicable to a given session, for which the firewall handles each incoming statement as follows:

  • If there is no applicable profile, there are no restrictions. The firewall accepts the statement.

  • If there is an applicable profile, its mode determines statement handling:

    • In RECORDING mode, the firewall adds the statement to the profile allowlist rules and accepts it.

    • In PROTECTING mode, the firewall compares the statement to the rules in the profile allowlist. The firewall accepts the statement if there is a match, and rejects it otherwise. If the mysql_firewall_trace system variable is enabled, the firewall also writes rejected statements to the error log.

    • In DETECTING mode, the firewall detects instrusions without denying access. The firewall accepts the statement, but also matches it to the profile allowlist, as in PROTECTING mode. If the statement is suspicious (nonmatching), the firewall writes it to the error log.

To use a firewall account profile to protect MySQL against incoming statements from a given account, follow these steps:

  1. Register the account profile and put it in RECORDING mode.

  2. Connect to the MySQL server using the account and execute statements to be learned. This trains the corresponding account profile and establishes the rules that form the profile allowlist.

  3. Switch the account profile to PROTECTING mode. When a client connects to the server using the account, the account profile allowlist restricts statement execution.

  4. Should additional training be necessary, switch the account profile to RECORDING mode again, update its allowlist with new statement patterns, then switch it back to PROTECTING mode.

By maintaining profiles, the firewall enables implementation of protection strategies such as these:

  • If an application has unique protection requirements, configure it to use an account not used for any other purpose and set up a corresponding account profile.

  • If related applications share protection requirements, configure them all to use the same account (and thus the same account profile).

Observe these guidelines for firewall-related account references:

  • Take note of the context in which account references occur. To name an account for firewall operations, specify it as a single quoted string ('user_name@host_name'). This differs from the usual MySQL convention for statements such as CREATE USER and GRANT, for which you quote the user and host parts of an account name separately ('user_name'@'host_name').

    The requirement for naming accounts as a single quoted string for firewall operations means that you cannot use accounts that have embedded @ characters in the user name.

  • The firewall assesses statements against accounts represented by actual user and host names as authenticated by the server. When registering account profiles, do not use wildcard characters or netmasks:

    • Suppose that an account named exists and a client uses it to connect to the server from the host

    • The account name contains a % wildcard character, but the server authenticates the client as having a user name of me and host name of, and that is what the firewall sees.

    • Consequently, the account name to use for firewall operations is rather than

The following example shows how to register an account profile with the firewall, teach the firewall the acceptable statements for that profile, and use the profile to protect MySQL against execution of unacceptable statements by the account. The example account, fwuser@localhost, is presumed for use by an application that accesses tables in the sakila database (available at

Use an administrative MySQL account to perform the steps in this procedure, except those steps designated for execution by the fwuser@localhost account corresponding to the account profile registered with the firewall. For statements executed using this account, the default database should be sakila. (You can use a different database by adjusting the instructions accordingly.)

  1. If necessary, create the account to use for executing statements (choose an appropriate password) and grant it privileges for the sakila database:

    CREATE USER 'fwuser'@'localhost' IDENTIFIED BY 'password';
    GRANT ALL ON sakila.* TO 'fwuser'@'localhost';
  2. Use the sp_set_firewall_mode() stored procedure to register an account profile with the firewall and place the profile in RECORDING (training) mode:

    CALL mysql.sp_set_firewall_mode('fwuser@localhost', 'RECORDING');

    As it executes, the stored procedure invokes firewall user-defined functions, which may produce output of their own.

  3. To train the registered account profile, connect to the server as fwuser from the server host so that the firewall sees a session account of fwuser@localhost. Then use the account to execute some statements to be considered legitimate for the profile. For example:

    SELECT first_name, last_name FROM customer WHERE customer_id = 1;
    UPDATE rental SET return_date = NOW() WHERE rental_id = 1;
    SELECT get_customer_balance(1, NOW());

    Because the profile is in RECORDING mode, the firewall records the normalized digest form of the statements as rules in the profile allowlist.


    Until the fwuser@localhost account profile receives statements in RECORDING mode, its allowlist is empty, which is equivalent to deny all. No statement can match an empty allowlist, which has these implications:

    • The account profile cannot be switched to PROTECTING mode. It would reject every statement, effectively prohibiting the account from executing any statement.

    • The account profile can be switched to DETECTING mode. In this case, the profile accepts every statement but logs it as suspicious.

  4. At this point, the account profile information is cached, as can be seen by querying the firewall INFORMATION_SCHEMA tables:

           WHERE USERHOST = 'fwuser@localhost';
    | MODE      |
           WHERE USERHOST = 'fwuser@localhost';
    | RULE                                                                       |
    | SELECT `first_name` , `last_name` FROM `customer` WHERE `customer_id` = ?  |
    | SELECT `get_customer_balance` ( ? , NOW ( ) )                              |
    | UPDATE `rental` SET `return_date` = NOW ( ) WHERE `rental_id` = ?          |
    | SELECT @@`version_comment` LIMIT ?                                         |

    The @@version_comment rule comes from a statement sent automatically by the mysql client when you connect to the server using the account corresponding to the account profile.


    Train the firewall under conditions matching application use. For example, to determine server characteristics and capabilities, a given MySQL connector might send statements to the server at the beginning of each session. If an application normally is used through that connector, train the firewall using the connector, too. That enables those initial statements to become part of the allowlist for the account profile associated with the application.

  5. Invoke sp_set_firewall_mode() again, this time switching the account profile to PROTECTING mode:

    CALL mysql.sp_set_firewall_mode('fwuser@localhost', 'PROTECTING');

    Switching the account profile out of RECORDING mode synchronizes its cached data to the mysql system database tables that provide persistent underlying storage. If you do not switch the mode for a profile that is being recorded, the cached data is not written to persistent storage and is lost when the server is restarted.

  6. Test the account profile by using the account to execute some acceptable and unacceptable statements. The firewall matches each statement against the profile allowlist and accepts or rejects it:

    • This statement is not identical to a training statement but produces the same normalized statement as one of them, so the firewall accepts it:

      mysql> SELECT first_name, last_name FROM customer WHERE customer_id = '48';
      | first_name | last_name |
      | ANN        | EVANS     |
    • These statements match nothing in the allowlist, so the firewall rejects each with an error:

      mysql> SELECT first_name, last_name FROM customer WHERE customer_id = 1 OR TRUE;
      ERROR 1045 (28000): Statement was blocked by Firewall
      mysql> SHOW TABLES LIKE 'customer%';
      ERROR 1045 (28000): Statement was blocked by Firewall
      mysql> TRUNCATE TABLE mysql.slow_log;
      ERROR 1045 (28000): Statement was blocked by Firewall
    • If the mysql_firewall_trace system variable is enabled, the firewall also writes rejected statements to the error log. For example:

      [Note] Plugin MYSQL_FIREWALL reported:
      'ACCESS DENIED for fwuser@localhost. Reason: No match in whitelist.
      Statement: TRUNCATE TABLE `mysql` . `slow_log` '

      These log messages may be helpful in identifying the source of attacks, should that be necessary.

The firewall account profile now is trained for the fwuser@localhost account. When clients connect using that account and attempt to execute statements, the profile protects MySQL against statements not matched by the profile allowlist.

It is also possible to detect intrusions by logging nonmatching statements as suspicious without denying access. First, put the account profile in DETECTING mode:

CALL mysql.sp_set_firewall_mode('fwuser@localhost', 'DETECTING');

Then, using the account, execute a statement that does not match the account profile allowlist. In DETECTING mode, the firewall permits the nonmatching statement to execute:

mysql> SHOW TABLES LIKE 'customer%';
| Tables_in_sakila (customer%) |
| customer                     |
| customer_list                |

In addition, the firewall writes a message to the error log:

[Note] Plugin MYSQL_FIREWALL reported:
'SUSPICIOUS STATEMENT from 'fwuser@localhost'. Reason: No match in whitelist.
Statement: SHOW TABLES LIKE ? '

DETECTING mode writes messages as Notes, which are information messages. To ensure that such messages appear in the error log and are not discarded, make sure that error-logging verbosity is sufficient to log information messages. For example, if you are using priority-based log filtering, as described in Priority-Based Error Log Filtering (log_filter_internal), set the log_error_verbosity system variable to a value of 3.

To disable an account profile, change its mode to OFF:

CALL mysql.sp_set_firewall_mode(user, 'OFF');

To forget all training for a profile and disable it, reset it:

CALL mysql.sp_set_firewall_mode(user, 'RESET');

The reset operation causes the firewall to delete all rules for the profile and set its mode to OFF.

Monitoring the Firewall

To assess firewall activity, examine its status variables. For example, after performing the procedure shown earlier to train and protect the fwuser@localhost account, the variables look like this:

mysql> SHOW GLOBAL STATUS LIKE 'Firewall%';
| Variable_name              | Value |
| Firewall_access_denied     | 3     |
| Firewall_access_granted    | 4     |
| Firewall_access_suspicious | 1     |
| Firewall_cached_entries    | 4     |

The variables indicate the number of statements rejected, accepted, logged as suspicious, and added to the cache, respectively. The Firewall_access_granted count is 4 because of the @@version_comment statement sent by the mysql client each of the three times you connected using the registered account, plus the SHOW TABLES statement that was not blocked in DETECTING mode.