Re: [SECURITY] DoS attack on backend possible - Mailing list pgsql-hackers

From ngpg@grymmjack.com
Subject Re: [SECURITY] DoS attack on backend possible
Date
Msg-id Xns9271B17F97FA9wn7t0983uom3iu23n@64.49.215.80
Whole thread Raw
In response to Re: [SECURITY] DoS attack on backend possible  (Florian Weimer <Weimer@CERT.Uni-Stuttgart.DE>)
List pgsql-hackers
Weimer@CERT.Uni-Stuttgart.DE (Florian Weimer) wrote 

> ngpg@grymmjack.com writes:
> 
>> if you are going to be passing any user input to the database, you 
>> must/should validate in some manner before blindly passing it to the db.
>> The db can and should guarantee data integrity, but the database cannot 
>> read your mind when it comes to how you structure your queries.
> 
> [example of SQL injection attack deleted]
> 
> This is not the problem at hand.  SQL injection attacks can be avoided
> easily.  Bugs in the conversion of strings to internal PostgreSQL
> objects are a different matter, though, and usually, devastating
> effects cannot be avoided by (reasonably complex) checks in the
> frontend.
> 

yeah i wasnt aware that adding a if(strlen($input) > SOME_REASONABLE_MAX) 
was complex.  the sql injection attack was just an(other) example of why 
you do not simply forward user input to the backend.  all i was trying to 
point out is that most of these buffer overflows in the backend can be 
avoided just as easily as the sql injection attack.


pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: Large file support available
Next
From: Bruce Momjian
Date:
Subject: Re: Build failure in current CVS