WebApp Sec mailing list archives

Re: Database Encryption -- Sql Injection


From: "Kevin Spett" <kspett () spidynamics com>
Date: Thu, 24 Apr 2003 10:52:58 -0400

I do agree that keeping it in the application is the best choice.  However,
it should be noted that if the crypto is being used correctly, ultimately,
it doesn't matter too much.  The security should rely on the secrecy of the
key.  In a well-implemented system, an attacker should be able to have the
algorithm and still not have a chance at deciphering the data.  You should
be using an industry standard cipher and not rolling your own.


Kevin Spett
SPI Labs
http://www.spidynamics.com/

----- Original Message -----
From: "Logan F.D. Greenlee" <lgreenlee () ciretose net>
To: "Dave Bergert" <dbergert () nobel-net com>; <webappsec () securityfocus com>
Sent: Monday, April 21, 2003 10:12 PM
Subject: RE: Database Encryption -- Sql Injection


Dave,
Your best bet will be to do unencryption in the application. The rationale
behind this is that if your database is compromised by a SQL injection
attack, all of the functions, sprocs, views and tables will be accessible by
the attacker. Thus, the decrypt function in the database can be utilized by
the attacker in compromising your data.
If a successful injection attack is executed against your database, the data
returned by the DB will be useless to the attacker unless they also have
access to the decrypt function in your application. By separating the two,
the attacker must succesfully compromise your application and your database
in order to compromise your data.

Logan

-----Original Message-----
From: Dave Bergert [mailto:dbergert () nobel-net com]
Sent: Monday, April 21, 2003 9:32 PM
To: webappsec () securityfocus com
Subject: Database Encryption -- Sql Injection


Does any one have any comments on where best to incorporate Column level
encryption in a Database field?  At the Database Server level (via a
User Defined Function) or at the Application Level. Which would be less
impervious to SQL Injection?

I am on a MS-SQL 2000 and IIS Platform.

If I had a User Defined Function for example:
 Select decrypt(AccountNumber, "key")  from tblTable where User =
'someuser'

If SQL Injection occurs:
 Select decrypt(AccountNumber, "key")  from tblTable where User =
'someuser' or 1=1

In this case if SQL injection occurs the encrypted field will be
automatically decrypted by the UDF... Showing all accountNumbers...


If I had the Decryption handled at the Application:
 Select encryptedAccountNumber from tblTable where User = 'someuser'

And had the application call:
 AccountNumber = DecryptFunction (ResultSet ("encryptedAccountNumber" ),
"key")


If SQL Injection occurs, the only way data could be seen if through
whatever mechanism the application displays the AccountNumber

(Are these scenarios identical ?)

I know that encryption is not a substitution for good input sanity
validation.
Which method would be better to implement?
Thanks for comments.


Regards,
Dave Bergert




Current thread: