I'm doing a CTF application for my final year project. I wanted to craft a SQL injection flag but I don't wish that they can simply found out the SQLi vulnerability by injecting ' . Hence, is there way to make SQL injection harder but not impossible using PHP?

  • What exactly do you want to achieve? I can't imagine what you need when saying harder and impossible... – Akam Apr 23 '13 at 11:57
  • Pretty much any SQL injection exercise can be both fun and educational, if you're required to determine something else first with another SQL injection, before winning the flag. This something else can be the data structure of same, or even other tables that are unknown but required for the final SQL injection. Or it could be getting data located in another record first, moving the record cursor up and down with previous injections. Or injecting new fields, or bogus records that would trigger stored procedures that aren't sanitized,... Options are only limited by your imagination really. – TildalWave Apr 23 '13 at 12:08

Blind SQL injection is probably one of the harder techniques, in that it requires more work. You could make a page that doesn't display any output based on the query, but shows error messages when a query is malformed. The attacker would then be forced to use blind SQLi techniques to attack it.

Of course, an automated tool would break this immediately, since it could automate the entire injection. One interesting way around that would be to have some kind of super-simple image CAPTCHA (no distortion, just a plain image) that the attacker must solve in order to send a query. This would either force them to do it manually, or make them look into OCR techniques.

  • Just curious but would this still be considered a Blind SQL Injection if error messages are still shown instead of generic ones? Most of the stuff that I've read say that a SQL injection is classified as blind when a generic error message (instead of database errors) is shown and the application doesn't allow dumping tables and stuff to the website itself. I'm currently learning about blind SQL injection and I'm confused how blind injection is classified. If you could shed some light on this, that would be awesome! Thanks. – CoderPE Nov 18 '18 at 2:00
  • 1
    @CoderPE Yes, it's still blind, because you're not directly getting any data out. All blind SQL injection requires some sort of oracle as a side-channel, whether that be an error or a time delay. Errors don't themselves leak table data (unless you emit a custom error that does) but instead tell you whether or not your query was valid - this is useful for building the initial injection without seeing the source. You can then leverage the errors as a side channel using lazy evaluation. – Polynomial Nov 18 '18 at 11:47

You could hide error details, which makes it harder for the user to know that injection is possible.

A next level would be Blind SQL injections, as Polynomial mentioned.

Also, you could reflect the injection on another page than the current page, for instance a log page so the user doesn't get imediate feedback on the injected code.

Or you could replace certain characters in the input string (See: SQL injection: isn't replace(“'”, “''”) good enough?).


My first thought is to limit what characters or regex patterns you look for. You could limit the use of only certain strings within the user input that would allow for attacks but would complicate what kind of attacks would work.

I'd say the easiest way to come up with something like this is probably to think of a particular attack you want someone to arrive at and then work backwards. Try to design protection to block everything other than the attack you wish them to perform. It could even be as simple as making it SQL Injection proof but then specifically allowing the desired attack input, though that might be a little too tricky depending on what you are looking for.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.