|Oracle8i Administrator's Guide
This chapter provides guidelines for developing security policies for database operation, and includes the following topics:
This section describes aspects of system security policy, and includes the following topics:
Each database has one or more administrators who are responsible for maintaining all aspects of the security policy: the security administrators. If the database system is small, the database administrator may have the responsibilities of the security administrator. However, if the database system is large, a special person or group of people may have responsibilities limited to those of a security administrator.
After deciding who will manage the security of the system, a security policy must be developed for every database. A database's security policy should include several sub-policies, as explained in the following sections.
Database users are the access paths to the information in an Oracle database. Therefore, tight security should be maintained for the management of database users. Depending on the size of a database system and the amount of work required to manage database users, the security administrator may be the only user with the privileges required to create, alter, or drop database users. On the other hand, there may be a number of administrators with privileges to manage database users. Regardless, only trusted individuals should have the powerful privileges to administer database users.
Database users can be authenticated (verified as the correct person) by Oracle using the host operating system, network services, or the database. Generally, user authentication via the host operating system is preferred for the following reasons:
User authentication by the database is normally used when the host operating system cannot support user authentication.
See Also: For more information about network authentication, see Oracle8i Distributed Database Systems.
For more information about user authentication, see "Creating Users".
If applicable, the following security issues must also be considered for the operating system environment executing Oracle and any database applications:
See Also: For more information about operating system security issues for Oracle databases, see your operating system-specific Oracle documentation.
Data security includes the mechanisms that control the access to and use of the database at the object level. Fine-grained access control can also limit data access to a more granular level. Your data security policy determines which users have access to a specific schema object, and the specific types of actions allowed for each user on the object. For example, user SCOTT can issue SELECT and INSERT statements but not DELETE statements using the EMP table. Your data security policy should also define the actions, if any, that are audited for each schema object.
Your data security policy will be determined primarily by the level of security you wish to establish for the data in your database. For example, it may be acceptable to have little data security in a database when you wish to allow any user to create any schema object, or grant access privileges for their objects to any other user of the system. Alternatively, it might be necessary for data security to be very controlled when you wish to make a database or security administrator the only person with the privileges to create objects and grant access privileges for objects to roles and users.
Overall data security should be based on the sensitivity of data. If information is not sensitive, then the data security policy can be more lax. However, if data is sensitive, a security policy should be developed to maintain tight control over access to objects.
This section describes aspects of user security policy, and includes the following topics:
For all types of database users, consider the following general user security issues:
If user authentication is managed by the database, security administrators should develop a password security policy to maintain database access security. For example, database users should be required to change their passwords at regular intervals, and of course, when their passwords are revealed to others. By forcing a user to modify passwords in such situations, unauthorized database access can be reduced.
To better protect the confidentiality of your password, Oracle can be configured to use encrypted passwords for client/server and server/server connections.
By setting the following values, you can require that the password used to verify a connection always be encrypted:
If enabled at both the client and server, passwords will not be sent across the network "in the clear", but will be encrypted using a modified DES (Data Encryption Standard) algorithm.
The DBLINK_ENCRYPT_LOGIN parameter is used for connections between two Oracle servers (for example, when performing distributed queries). If you are connecting from a client, Oracle checks the ORA_ENCRYPT_LOGIN environment variable.
Whenever you attempt to connect to a server using a password, Oracle encrypts the password before sending it to the server. If the connection fails and auditing is enabled, the failure is noted in the audit log. Oracle then checks the appropriate DBLINK_ENCRYPT_LOGIN or ORA_ENCRYPT_LOGIN value. If it set to FALSE, Oracle attempts the connection again using an unencrypted version of the password. If the connection is successful, the connection replaces the previous failure in the audit log, and the connection proceeds. To prevent malicious users from forcing Oracle to re-attempt a connection with an unencrypted version of the password, you must set the appropriate values to TRUE.
Security administrators should consider issues related to privilege management for all types of users. For example, in a database with many usernames, it may be beneficial to use roles (which are named groups of related privileges that you grant to users or other roles) to manage the privileges available to users. Alternatively, in a database with a handful of usernames, it may be easier to grant privileges explicitly to users and avoid the use of roles.
Security administrators managing a database with many users, applications, or objects should take advantage of the benefits offered by roles. Roles greatly simplify the task of privilege management in complicated environments.
Security administrators must also define a policy for end-user security. If a database is large with many users, the security administrator can decide what groups of users can be categorized, create user roles for these user groups, grant the necessary privileges or application roles to each user role, and assign the user roles to the users. To account for exceptions, the security administrator must also decide what privileges must be explicitly granted to individual users.
Roles are the easiest way to grant and manage the common privileges needed by different groups of database users.
Consider a situation where every user in the accounting department of a company needs the privileges to run the ACCTS_RECEIVABLE and ACCTS_PAYABLE database applications. Roles are associated with both applications, and contain the object privileges necessary to execute those applications.
The following actions, performed by the database or security administrator, address this simple security situation:
This security model is illustrated in Figure 22-1.
This plan addresses the following potential situations:
When possible, utilize roles in all possible situations to make end-user privilege management efficient and simple.
Security administrators should have a policy addressing administrator security. For example, when the database is large and there are several types of database administrators, the security administrator may decide to group related administrative privileges into several administrative roles. The administrative roles can then be granted to appropriate administrator users. Alternatively, when the database is small and has only a few administrators, it may be more convenient to create one administrative role and grant it to all administrators.
After database creation, immediately change the passwords for the administrative SYS and SYSTEM usernames to prevent unauthorized access to the database. Connecting as SYS and SYSTEM gives a user the powerful privileges to modify a database in many ways. Therefore, privileges for these usernames are extremely sensitive, and should only be available to select database administrators.
See Also: The passwords for these accounts can be modified using the procedures described in "Altering Users".
Only database administrators should have the capability to connect to a database with administrator privileges. Connecting as SYSDBA gives a user unrestricted privileges to do anything to a database (such as startup, shutdown, and recover) or the objects within a database (such as create, drop, and delete from). Only users with SYS-privileged connections can alter data dictionary tables (for example,
connect as SYSDBA/SYSOPER).
Roles are the easiest way to restrict the powerful system privileges and roles required by personnel administrating of the database.
Consider a scenario where the database administrator responsibilities at a large installation are shared among several database administrators, each responsible for the following specific database management jobs:
In this scenario, the security administrator should structure the security for administrative personnel as follows:
This plan diminishes the likelihood of future problems in the following ways:
Security administrators must define a special security policy for the application developers using a database. A security administrator may grant the privileges to create necessary objects to application developers. Alternatively, the privileges to create objects may only be granted to a database administrator, who receives requests for object creation from developers.
Database application developers are unique database users who require special groups of privileges to accomplish their jobs. Unlike end users, developers need system privileges, such as CREATE TABLE, CREATE PROCEDURE, and so on. However, only specific system privileges should be granted to developers to restrict their overall capabilities in the database.
In many cases, application development is restricted to test databases and not allowed on production databases. This restriction ensures that application developers do not compete with end users for database resources, and that they cannot detrimentally affect a production database.
After an application has been thoroughly developed and tested, it is permitted access to the production database and made available to the appropriate end users of the production database.
The database administrator can define the following options when determining which privileges should be granted to application developers:
An application developer is allowed to create new schema objects, including tables, indexes, procedures, packages, and so on. This option allows the application developer to develop an application independent of other objects.
An application developer is not allowed to create new schema objects. All required tables, indexes, procedures, and so on are created by a database administrator, as requested by an application developer. This option allows the database administrator to completely control a database's space usage and the access paths to information in the database.
Although some database systems use only one of these options, other systems could mix them. For example, application developers can be allowed to create new stored procedures and packages, but not allowed to create tables or indexes. A security administrator's decision regarding this issue should be based on the following:
Security administrators can create roles to manage the privileges required by the typical application developer. For example, a typical role named APPLICATION_DEVELOPER might include the CREATE TABLE, CREATE VIEW, and CREATE PROCEDURE system privileges. Consider the following when defining roles for application developers:
While application developers are typically given the privileges to create objects as part of the development process, security administrators must maintain limits on what and how much database space can be used by each application developer. For example, as the security administrator, you should specifically set or restrict the following limits for each application developer:
See Also: Both limitations can be set by altering a developer's security domain. For more information, see "Altering Users".
In large database systems with many database applications (for example, precompiler and Forms applications), you might want to have application administrators. An application administrator is responsible for the following types of tasks:
Often, an application administrator is also the application developer who designed the application. However, these jobs might not be the responsibility of the developer and can be assigned to another individual familiar with the database application.
Database security systems depend on passwords being kept secret at all times. Still, passwords are vulnerable to theft, forgery, and misuse.To allow for greater control over database security, Oracle's password management policy is controlled by DBAs.
This section describes the following aspects of Oracle password management:
When a particular user exceeds a designated number of failed login attempts, the server automatically locks that user's account. DBAs specify the permissible number of failed login attempts using the CREATE PROFILE statement. DBAs also specify the amount of time accounts remain locked.
In the following example, the maximum number of failed login attempts for the user ASHWINI is 4, and the amount of time the account will remain locked is 30 days; the account will unlock automatically after the passage of 30 days.
CREATE PROFILE prof LIMIT FAILED_LOGIN_ATTEMPTS 4 PASSWORD_LOCK_TIME 30; ALTER USER ashwini PROFILE prof;
If the DBA does not specify a time interval for unlocking the account, ACCOUNT_LOCK _TIME reverts to a default value. If the DBA specifies ACCOUNT_LOCK_TIME as UNLIMITED, then the system security officer must explicitly unlock the account. Thus, the amount of time an account remains locked depends upon how the DBA configures the resource profile assigned to the user.
After a user successfully logs into an account, that user's unsuccessful login attempt count, if there is one, is reset to 0.
The security officer can also explicitly lock user accounts. When this occurs, the account cannot be unlocked automatically; only the security officer should unlock the account.
See Also: For more information about the CREATE PROFILE statement, see the Oracle8i SQL Reference.
DBAs use the CREATE PROFILE statement to specify a maximum lifetime for passwords. When the specified amount of time passes and the password expires, the user or DBA must change the password. The following statement indicates that ASHWINI can use the same password for 90 days before it expires:
CREATE PROFILE prof LIMIT FAILED_LOGIN_ATTEMPTS 4 PASSWORD_LOCK_TIME 30 PASSWORD_LIFE_TIME 90; ALTER USER ashwini PROFILE prof;
DBAs can also specify a grace period using the CREATE PROFILE statement. Users enter the grace period upon the first attempt to login to a database account after their password has expired. During the grace period, a warning message appears each time users try to log in to their accounts, and continues to appear until the grace period expires. Users must change the password within the grace period. If the password is not changed within the grace period, the account expires and no further logins to that account are allowed until the password is changed. Figure 22-2 shows the chronology of the password lifetime and grace period.
For example, the lifetime of a password is 60 days, and the grace period is 3 days. If the user tries to log in on any day after the 60th day (this could be the 70th day, 100th day, or another; the point here is that it is the first login attempt after the password lifetime), that user receives a warning message indicating that the password is about to expire in 3 days. If the user does not change the password within three days from the first day of the grace period, the user's account expires. The following statement indicates that the user must change the password within 3 days of its expiration:
CREATE PROFILE prof LIMIT FAILED_LOGIN_ATTEMPTS 4 ACCOUNT_LOCK_TIME 30 PASSWORD_GRACE_TIME 3; ALTER USER ashwini PROFILE prof;
The security officer can also explicitly expire the account. This is particularly useful for new accounts.
See Also: For more information about the CREATE PROFILE statement, see Oracle8i SQL Reference.
DBAs use the CREATE PROFILE statement to specify a time interval during which users cannot reuse a password.
In the following statement, the DBA indicates that the user cannot reuse her password for 60 days.
CREATE PROFILE prof LIMIT PASSWORD_REUSE_TIME 60 PASSWORD_REUSE_MAX UNLIMITED;
The next statement shows that the number of password changes the user must make before her current password can be used again is 3.
CREATE PROFILE prof LIMIT PASSWORD_REUSE_MAX 3 PASSWORD_REUSE_TIME UNLIMITED;
If you specify PASSWORD_REUSE_TIME or PASSWORD_REUSE_MAX, you must set the other to UNLIMITED or not specify it at all.
Oracle's password complexity verification routine can be specified using a PL/SQL script (utlpwdmg.sql), which sets the default profile parameters.
The password complexity verification routine performs the following checks:
Oracle recommends that you do not change passwords using the ALTER USER statement because it does not fully support the password verification function. Instead, you should use OCIPasswordChange() to change passwords.
DBAs can enhance the existing password verification complexity routine or create their own password verification routines using PL/SQL or third-party tools.
The DBA-authored PL/SQL call must adhere to the following format:
password_parameterIN VARCHAR (30), old_password_parameter IN VARCHAR (30) ) RETURN BOOLEAN
After a new routine is created, it must be assigned as the password verification routine using the user's profile or the system default profile.
The password verify routine must be owned by SYS.
The following sample script sets default password resource limits and provides minimum checking of password complexity. You can use this sample script as a model when developing your own complexity checks for a new password.
This script sets the default password resource parameters, and must be run to enable the password features. However, you can change the default resource parameters if necessary.
The default password complexity function performs the following minimum complexity checks:
This function must be created in SYS schema, and you must
sys/<password> as sysdba before running the script.
CREATE OR REPLACE FUNCTION verify_function (username varchar2, password varchar2, old_password varchar2) RETURN boolean IS n boolean; m integer; differ integer; isdigit boolean; ischar boolean; ispunct boolean; digitarray varchar2(20); punctarray varchar2(25); chararray varchar2(52); BEGIN digitarray:= '0123456789'; chararray:= 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'; punctarray:='!"#$%&()''*+,-/:;<=>?_'; --Check if the password is same as the username IF password = username THEN raise_application_error(-20001, 'Password same as user'); END IF; --Check for the minimum length of the password IF length(password) < 4 THEN raise_application_error(-20002, 'Password length less than 4'); END IF; --Check if the password is too simple. A dictionary of words may be --maintained and a check may be made so as not to allow the words --that are too simple for the password. IF NLS_LOWER(password) IN ('welcome', 'database', 'account', 'user', 'password', 'oracle', 'computer', 'abcd') THEN raise_application_error(-20002, 'Password too simple');
END IF; --Check if the password contains at least one letter, one digit and one
--punctuation mark. --1. Check for the digit --You may delete 1. and replace with 2. or 3. isdigit:=FALSE; m := length(password); FOR i IN 1..10 LOOP FOR j IN 1..m LOOP IF substr(password,j,1) = substr(digitarray,i,1) THEN isdigit:=TRUE; GOTO findchar; END IF; END LOOP; END LOOP; IF isdigit = FALSE THEN raise_application_error(-20003, 'Password should contain at least one digit, one character and one punctuation'); END IF; --2. Check for the character <<findchar>> ischar:=FALSE; FOR i IN 1..length(chararray) LOOP FOR j IN 1..m LOOP IF substr(password,j,1) = substr(chararray,i,1) THEN ischar:=TRUE; GOTO findpunct; END IF; END LOOP; END LOOP; IF ischar = FALSE THEN raise_application_error(-20003, 'Password should contain at least one digit, one character and one punctuation'); END IF; --3. Check for the punctuation <<findpunct>> ispunct:=FALSE; FOR i IN 1..length(punctarray) LOOP FOR j IN 1..m LOOP IF substr(password,j,1) = substr(punctarray,i,1) THEN ispunct:=TRUE; GOTO endsearch; END IF; END LOOP; END LOOP; IF ispunct = FALSE THEN raise_application_error(-20003, 'Password should contain at least one \ digit, one character and one punctuation'); END IF; <<endsearch>> --Check if the password differs from the previous password by at least 3 letters IF old_password = '' THEN raise_application_error(-20004, 'Old password is null'); END IF; --Everything is fine; return TRUE ; differ := length(old_password) - length(password); IF abs(differ) < 3 THEN IF length(password) < length(old_password) THEN m := length(password); ELSE m:= length(old_password); END IF; differ := abs(differ); FOR i IN 1..m LOOP IF substr(password,i,1) != substr(old_password,i,1) THEN differ := differ + 1; END IF; END LOOP; IF differ < 3 THEN raise_application_error(-20004, 'Password should differ by at least 3 characters'); END IF; END IF; --Everything is fine; return TRUE ; RETURN(TRUE); END;
Security administrators should define a policy for the auditing procedures of each database. You may, for example, decide to have database auditing disabled unless questionable activities are suspected. When auditing is required, the security administrator must decide what level of detail to audit the database; usually, general system auditing is followed by more specific types of auditing after the origins of suspicious activity are determined.