Oracle PL/SQL for DBAs

Specifying Audit Columns

If we record information every time someone selects anything from a table, the audit trail will become very large, making it difficult to manage. You may want to limit recording of accesses to only a specific set of columns. Let’s revisit the description of the table EMP.

    SQL> DESC emp
     Name              Null?    Type
     ----------------- -------- ------------
     EMPID             NOT NULL NUMBER(4)
     EMPNAME                    VARCHAR2(10)
     JOB                        VARCHAR2(9)
     MGR                        NUMBER(4)
     HIREDATE                   DATE
     SALARY                     NUMBER(7,2)
     COMM                       NUMBER(7,2)
     DEPTNO                     NUMBER(2)

Examining the columns, you’ll notice that some columns may be considered more important to audit than others. For example, you may want to make sure that all accesses to the column SALARY are logged, but you might not want to audit such columns as HIREDATE quite so stringently. In this example, let’s assume that I want to audit accesses only to SALARY and COMM, not to all columns. I can do so by specifying a value for the ADD_POLICY procedure’s audit_column parameter as follows.

    BEGIN
       DBMS_FGA.add_policy (object_schema      => 'HR',
                            object_name        => 'EMP',
                            policy_name        => 'EMP_SEL',
                            
                            audit_column       => 'SALARY, COMM'
                           );
    END;

This causes the trail to be generated only if the user selects the SALARY or COMM columns. If she selects only ENAME, the trail is not recorded.

What I’ve described applies not only to columns named explicitly in the query, but also to columns referenced implicitly. For instance the query:

    SELECT * FROM hr.emp;

selects all columns from the table EMP, including COMM and SALARY. ...