In Postgres, are prepared queries and user defined functions equivalent as a mechanism for guarding against SQL injection?
Are there particular advantages in one approach over the other?
1 Answer
It depends.
Prepared statements
A.k.a. "prepared queries". These only take values when executed with EXECUTE
. No SQL-injection possible at this stage. But you have to defend against SQL-injection while generating / concatenating the statement to be fed to PREPARE
in the first place. Prepared statements are very much equivalent to stored SQL functions / procedures in this regard.
SQL functions / procedures
With LANGUAGE sql
they are generally safe.
Passed parameters are treated as values and SQL-injection is not possible - as long as you don't call unsafe functions from the body and pass parameters.
PL/pgSQL functions /procedures
With LANGUAGE plpgsql
they are normally safe.
However, PL/pgSQL allows for dynamic SQL where passed parameters (or parts) are concatenated to a query string and executed with EXECUTE
. This can convert user input to SQL code and make SQL injection possible. Tools are there to do it safely. You cannot tell from outside whether the function body handles it properly, you have to look at the code.
Plain SQL statements using parameters as values are safe against SQL injection, just like SQL functions. Only use dynamic SQL where needed and follow these guidelines:
Preferably pass values as values to EXECUTE
with the USING
clause. Makes SQL injection impossible on principal. Example.
Never just build strings from user input and execute. This includes identifiers, directly passed by a user or fetched from a system catalog.
When concatenating values in the SQL string, wrap 'strings' in single-quotes safely. Use:
When concatenating identifiers in the SQL string, wrap them in double-quotes safely where required. Use:
format()
with format specifier%I
. Example.quote_ident()
. Example.- a cast to a registered type -
regclass
for table names:_tbl::regclass
. (Only works for existing objects!) Example.
Concatenating other parts of the SQL code is never safe, when allowing unchecked user input in any way.
Related:
- Refactor a PL/pgSQL function to return the output of various SELECT queries
- Define table and column names as arguments in a plpgsql function?
More about performance implications in this related answer:
Basics on SQL-injection:
Similar considerations apply to other server-side languages that allow dynamic SQL.
-
So, in summary: If 1) I use only language sql, I'm safe, 2) if I use plpgslq but not execute, I'm safe, 3) if I use plpgsql and execute but no identifiers and %s or %L as appropriate I'm safe, or 4) if I use plpgsql and execute and identifiers but %I or quote_ident as appropriate I'm safe. Correct?– user4930Commented Sep 12, 2013 at 17:28
-
@mickeyf: Basically yes. Plus, use the
USING
clause for passing values toEXECUTE
whenever possible. You could call a PL/pgSQL function from within an SQL function and pass parameters. So, to be absolutely correct, you are safe as long as you don't call any unsafe functions directly or indirectly. If all your functions are done properly, that cannot happen. Commented Sep 12, 2013 at 21:13 -
Would "Concatenating other parts of the SQL code" be safe if done in
FORMAT
with%s
?– RafsCommented Nov 30, 2023 at 12:52 -
@Rafs Generally no. There are exceptions, where SQLi is not possible a priori. Commented Nov 30, 2023 at 15:00