Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:108537
MIME-Version: 1.0
References: <CAFv4g+EG6f9e6yJqp1sesbJstYJwpdkVso3CKjAD_EAkcrtHqA@mail.gmail.com>
In-Reply-To: <CAFv4g+EG6f9e6yJqp1sesbJstYJwpdkVso3CKjAD_EAkcrtHqA@mail.gmail.com>
Date: Thu, 13 Feb 2020 12:31:44 +0000
Message-ID: <CAFv4g+Fa_cnyWqEP-qo9dgBiz7qszK-s52GuxOHfF6k1GmFcDQ@mail.gmail.com>
To: PHP internals <internals@lists.php.net>
Content-Type: multipart/alternative; boundary="0000000000003ca9ce059e74466d"
Subject: Re: Literal / Taint checking
From: craig@craigfrancis.co.uk (Craig Francis)

--0000000000003ca9ce059e74466d
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi,

While there was a brief discussion about an *is_literal*() method in
August, I'm wondering where I can go next?

Just as a reminder, the main objection seemed to be that Taint checking is
the current solution. For example, those created by Laruence[1],
MediaWiki[2], and Matthew[3]. But this can never be as good at the PHP
engine explicitly stating a variable *only* contains literal values, where
it can be checked at runtime, and be a key part of the development process.

And while I'm using SQL injection in my examples (because it's easy to show
how it can enforce the use of parameterised queries); it would also be
useful to protect against command line injection, and HTML/XSS as well
(e.g. a templating system can only accept HTML as literal strings, and
the user supplied values be provided separately).

I'm assuming this would change the zval structure (to include an
"is_literal" flag?), and it would be more of a PHP 8.0 change, rather than
8.1.

Craig


---

Broken taint check, due to missing quote marks:

$sql =3D =E2=80=98... WHERE id =3D =E2=80=99 . mysqli_real_escape_string($d=
b, $_GET[=E2=80=98id=E2=80=99]);

---

Support for "WHERE ... IN", ideally done via an abstraction, so you don't
need to write this every time:

$sql =3D '... WHERE id IN (' . substr(str_repeat('?,', count($ids)), 0, -1)=
 .
')';

---

[1] https://github.com/laruence/taint
[2] https://www.mediawiki.org/wiki/Phan-taint-check-plugin
[3] https://psalm.dev/r/ebb9522fea

---




On Thu, 15 Aug 2019 at 19:02, Craig Francis <craig@craigfrancis.co.uk>
wrote:

> Hi,
>
> How likely would it be for PHP to do Literal tracking of variables?
>
> This is something that's being discussed JavaScript TC39 at the moment
> [1], and I think it would be even more useful in PHP.
>
> We already know we should use parameterized/prepared SQL, but there is no
> way to prove the SQL string hasn't been tainted by external data in large
> projects, or even in an ORM.
>
> This could also work for templating systems (blocking HTML injection) and
> commands.
>
> Internally it would need to introduce a flag on every variable, and a
> single function to check if a given variable has only been created by
> Literal(s).
>
> Unlike the taint extension, there should be no way to override this (e.g.
> no taint/untaint functions); and if it was part of the core language, it
> will continue to work after every update.
>
> One day certain functions (e.g. mysqli_query) might use this information
> to generate a error/warning/notice; but for now, having it available for
> checking would be more than enough.
>
> Craig
>
>
>
>     public function exec($sql, $parameters =3D []) {
>         if (!*is_literal*($sql)) {
>             throw new Exception('SQL must be a literal.');
>         }
>         $statement =3D $this->pdo->prepare($sql);
>         $statement->execute($parameters);
>         return $statement->fetchAll();
>     }
>
> ...
>
>     $sql =3D 'SELECT * FROM table WHERE id =3D ?';
>
>     $result =3D $db->exec($sql, [$id]);
>
>
>
> [1] https://github.com/tc39/proposal-array-is-template-object
> https://github.com/mikewest/tc39-proposal-literals
>

--0000000000003ca9ce059e74466d--