On 5 Oct 2011, at 02:02, Tommy Pham wrote:
> On Tue, Oct 4, 2011 at 5:51 PM, Stuart Dallas <stu...@3ft9.com> wrote:
> On 5 Oct 2011, at 01:13, Tommy Pham wrote:
>
>> On Tue, Oct 4, 2011 at 4:49 PM, Stuart Dallas <stu...@3ft9.com> wrote:
>>
>> On 5 Oct 2011, at 00:45, Tommy Pham wrote:
>>
>>> On Tue, Oct 4, 2011 at 4:11 PM, Stuart Dallas <stu...@3ft9.com> wrote:
>>> On 5 Oct 2011, at 00:04, Mark Kelly wrote:
>>>
>>> > Hi.
>>> >
>>> > On Tuesday 04 Oct 2011 at 21:39 Stuart Dallas wrote:
>>> >
>>> >> http://stut.net/2011/09/15/mysql-real-escape-string-is-not-enough/
>>> >
>>> > Thanks. I followed this link through and read the full message (having
>>> > missed
>>> > it the first time round), and while I find the idea of using base64 to
>>> > sanitise text interesting I can also forsee a few difficulties:
>>> >
>>> > It would prevent anyone from accessing the database directly and getting
>>> > meaningful results unless the en/decode is in triggers, or maybe stored
>>> > procedures. No more one-off command-line queries.
>>> >
>>> > How would you search an encoded column for matching text?
>>> >
>>> > I'd be interested in any ideas folk have about these issues, or any others
>>> > they can envisage with this proposal.
>>>
>>> Base64 encoding will work when the native base64 functions are available in
>>> MySQL which will allow you to base64 encode the data into a statement like
>>> INSERT INTO table SET field = FROM_BASE64("<?php echo base64_encode($data);
>>> ?>") sorta thing. I'm still not a massive fan of that idea given that
>>> prepared statements are an option, but it would work.
>>>
>>>
>>> Inserting and updating isn't the problem. I think Mark referring to is how
>>> would that be implemented in this simple type of query:
>>>
>>> SELECT * FROM my_table WHERE col_name LIKE '%key word%';
>>>
>>> If there's no viable mean to filter the data, that storage method/medium is
>>> rather pointless, IMHO.
>>
>> Go back and read what I wrote again. Base64 is only being used to transmit
>> the data to MySQL - it's being stored in the database in its decoded form.
>>
>>
>> The question still applies as how would you safeguard that 'key word'
>> transmission, especially against SQL injection. I suppose one could do it
>> this way:
>>
>> SELECT * FROM my_table WHERE col_name LIKE CONCAT('%', FROM_BASE64("<?php
>> echo base64_encode($data); ?>"), '%')
>>
>> Is the overhead worth it to warrant that kind of safeguard? That's just a
>> simple query with a simple search criteria. What about in the case of
>> subselect and multi-table joins?
>
> That would indeed be logical if base64 was your chosen method of protection,
> but I think prepared statements are a far more elegant solution. As for the
> overhead I very much doubt there's much difference between that and the
> overhead of prepared statements.
>
>
> IIRC, prepared statements doesn't incur any overhead. Instead, it's supposed
> to enhance performance by telling SQL to 'prepare' via compilation. So if
> you're comparing performance between the overhead of base64 vs prepared
> statement, then the difference would be quite clear, especially when the
> table(s) is/are more than a couple hundred thousand rows and the queri(es)
> are complex. This is not mention the added complexity into the application
> where managing and expanding it would incur real (developer time) overhead,
> IMO.
Prepared statements incur an additional hit against the DB server to prepare
the statement.
The cost of using base64 in the manner suggested is minimal, regardless of the
size of the data. The MySQL query analyser is intelligent enough to know that
from_base64('xyz') is a constant expression and will therefore only evaluate it
once.
As for the added complexity, if you have SQL statements all over your code then
yes it will add a time overhead, but any codebase of a significant size should
be using a centralised API for database access such that changes like this have
a very limited scope.
-Stuart
--
Stuart Dallas
3ft9 Ltd
http://3ft9.com/
