Ah, external developers
What a short-sighted money saver they are.
Database security expert David Litchfield has published details of a new type of database attack technique. Lateral SQL injection creates a means for hackers to access database data or inject hostile code onto vulnerable systems. Exploitation is difficult and only possible in limited circumstances, Litchfield notes. …
Sack the 'tard for not using bind variables in his sql code and get someone who knows what they are doing to do the work. It's not as if Tom Kyte hasn't been banging on about this since the beginning of the millenium.
Sick vulture because it could swallow a program spec and puke better code than that.
.. simply break the idiot's kneecaps with a well wielded lead pipe.
Not using bind variables is beyond idiotic.
Using DBMS_SQL, the most complex cursor interface in PL/SQL, is just plain stupid - the only time that DBMS_SQL is needed is when requiring a describe interface for a cursor.
And guess what - needing a describe interface means you have no fricken clue what the SQL projection is, thus dynamic SQL. And dynamic SQL is *ALWAYS* risky - and if you code that without a healthy dose of logic (and bind variables!), and then get hack.. your poor dumb bastard you, you deserved that.
What is sad though that these "lateral" SQL injections will likely work on many a system. Not because Oracle is that flawed. But because these code monkeys (not to be confused with real programmers) that write PL/SQL are just that damn stupid.
Seems like many is confusing the "unbreakable" slogan of Oracle to extend to their code monkeys. Sadly, that is not the case.
surely an exception section :
EXCEPTION WHEN OTHERS THEN
BEGIN
DBMS_SQL.CLOSE_CURSOR(CURSOR_NAME);
EXCEPTION WHEN OTHERS THEN RAISE;
END;
RAISE;
Error handling should always exist, regardless of whether you expect it to happen or not. I'm sure there's an easier, more elegant solution, but off the top of my head that should work. Cursors should always be closed in the event of an error!
As Dominic says this is VERY easy to avoid - simply binding the variable would prevent the attack - in addition it would likely save you resources as well as it would only soft parse the query instead of the hard parse everytime it has a new date plugged into the code.
@Aodhhan - is it a vulnerability though? Poor coding practice yes, but a vuln that needs fixing? I'm not convinced - IMHO it's a bit like saying that because you can shoot yourself in the foot with pointers in C, C needs fixing...