You have a database driver that takes care of communicating with the database.
In the bad old days (pre-early 2000s) the only way they knew how to do that was plain old SQL strings so you passed a string that contained both the data and the instructions on what to do with it.
Now you SHOULD be writing prepared statements that contain the instructions then passing the data separately to fill in the placeholders in the prepared statement via the driver (NOT via modifying the string).
// DO NOT DO THISexecute("INSERT INTO foo VALUES ('a', 'b', 'c')")
vs
// DO THISexecutePrepared("INSERT INTO foo VALUES (?,?,?)", "a", "b", "c")
Because it’s literally impossible for SQL injection to occur if you do this. The database has already compiled the operation. There’s nothing to escape, there’s no more logic that can be added, you’re free to insert arbitrary gook just like you can into any old array.
It’s a common problem for the same reason that it’s a common problem for people to have precision errors when doing math with currencies… People write the wrong code because they don’t know any better (in that case using float or double/floating point math instead of a BigDecimal type).
Not filtering out characters that could be part of URL has no bearing on whether or not the site is properly protected from SQL injection. I’m much more often worried about sites that explicitly filter out certain characters because it likely means they don’t understand what they’re doing (similar to sites that insist on annual password changes).
The fact that people are arguing about this shows how much of an issue we have with education on this topic.
That’s not how this works.
You have a database driver that takes care of communicating with the database.
In the bad old days (pre-early 2000s) the only way they knew how to do that was plain old SQL strings so you passed a string that contained both the data and the instructions on what to do with it.
Now you SHOULD be writing prepared statements that contain the instructions then passing the data separately to fill in the placeholders in the prepared statement via the driver (NOT via modifying the string).
// DO NOT DO THIS execute("INSERT INTO foo VALUES ('a', 'b', 'c')")
vs
// DO THIS executePrepared("INSERT INTO foo VALUES (?,?,?)", "a", "b", "c")
Please explain how you remain confident of that “SHOULD” when they are not sanitizing the HTML out?
Because it’s literally impossible for SQL injection to occur if you do this. The database has already compiled the operation. There’s nothing to escape, there’s no more logic that can be added, you’re free to insert arbitrary gook just like you can into any old array.
“if” caring a lot of water on this here frog’s back mr. scorpion.
And yet injection is still #3 in the OWASP Top 10
Okay?
You make it sound as if it’s a thing of the past when it is still a common problem.
It’s a common problem for the same reason that it’s a common problem for people to have precision errors when doing math with currencies… People write the wrong code because they don’t know any better (in that case using float or double/floating point math instead of a BigDecimal type).
Not filtering out characters that could be part of URL has no bearing on whether or not the site is properly protected from SQL injection. I’m much more often worried about sites that explicitly filter out certain characters because it likely means they don’t understand what they’re doing (similar to sites that insist on annual password changes).
The fact that people are arguing about this shows how much of an issue we have with education on this topic.