| blob | 2a2fdce1bf7b9e9bbc53e6db5e2811e7d0694e71 |
4 <head>
7 <meta name="keywords" content="HTMLPurifier, HTML Purifier, HTML, filter, filtering, HTML_Safe, PEAR, comparison, kses, striptags, SafeHTMLChecker" />
12 <!--[if lt IE 7.]><script defer="defer" type="text/javascript" src="./pngfix.js"></script><![endif]-->
14 </head>
15 <body>
20 <div id="header"><a href="./"><span class="html">HTML</span> <span class="purifier">Purifier</span></a></div>
26 gone from passive consumer to active producer of content on the World Wide
30 put the user in control.</p>
32 <p>Give the user too much control, however, and you set yourself up
36 has proven to be both a blessing and a curse, and the software that processes
37 it must strike a fine balance between security and usability. How do
38 we prevent users from injecting JavaScript or inserting malformed
40 a rich syntax of tags, attributes and <acronym
44 about sloppy coding messing up <acronym
47 developer has come across this problem before, and many have tried
48 (albeit unsuccessfully) to solve this problem. We will analyze existing
49 libraries to demonstrate how they are ineffective and, of course,
51 standards-compliance.</p>
53 <p>I will take no quarter and pull no punches: as of the time of writing,
55 for richly formatted documents. But, nonetheless, there is a necessary
56 disclaimer:</p>
59 This comparison document was written by the author of HTML Purifier,
61 mean that it is biased: I have made every attempt to be <strong>factual and
62 fair</strong>, and I hope that you will agree, by the time you finish reading
63 this document, that HTML Purifier is the only satisfactory HTML
64 filter out there today.
65 </p>
69 generated. Please enable JavaScript to see it.</p>
70 </noscript></div>
79 <thead>
80 <tr>
92 </tr>
93 </thead>
95 <tbody>
97 <tr>
109 </tr>
111 <tr>
123 </tr>
125 <tr>
137 </tr>
139 <tr>
151 </tr>
153 <tr>
165 </tr>
167 <tr>
179 </tr>
181 </tbody>
183 </table>
184 </div>
186 <p class="lead"><a href="#Tidy">HTML Tidy</a> is omitted from this list because it is not an HTML
187 filter.</p>
193 A clever person solves a problem.
194 A wise person avoids it.
195 </div>
197 </blockquote>
200 of HTML filters, we must first consider another domain: alternate
201 markup libraries. While libraries of this type really shouldn't be
203 they are the number one method of taking user input and processing it into
204 something more than plain old text. These libraries forgo
210 such markup languages (although it should be noted that
211 Wikitext and Markdown can allow
213 The benefits (to those who use it, anyway) are clear: simplicity and
214 security.
215 </p>
218 <thead>
219 <tr>
222 </tr>
223 </thead>
224 <tbody>
225 <tr>
228 </tr>
229 <tr>
232 </tr>
233 <tr>
236 </tr>
237 <tr>
240 </tr>
241 <tr>
243 <td><tt><b>B</b> <i>i</i> <a href="http://www.example.com/">link</a></tt></td>
244 </tr>
245 <tr>
248 </tr>
249 </tbody>
250 </table>
255 There are many variants of Wikitext currently extant.</li>
256 <li>Strictly speaking, the Markdown syntax is not equivalent: bold text
259 however, map those two semantic tags to the associated styling, so
260 many users assume that it really is italics (and use it improperly for,
261 say, book titles.)</li>
262 </ol>
267 source code is often criticized for being difficult to read. For example,
268 compare:</p>
270 <pre>
271 * Item 1
272 * Item 2
273 </pre>
277 <pre>
282 </pre>
284 <p>Which would you prefer to edit? The answer seems obvious, but be careful
285 not to fall into the fallacy of <a
289 editor, which blows earlier choices out of the water in terms
290 of usability.</p>
292 <p>Note that rich text editors and alternate markup syntaxes are not
293 mutually exclusive, but, when push comes to shove, it's easier
294 implement this sort of editor on top of <acronym
296 markup language. And in the cases when it is done, you usually end up with
297 a live preview, not a true rich text editor.</p>
302 editors aren't all that great." There are many good arguments
303 against these editors, and <a
305 people have written essays</a> devoted to
307 In addition to the usual arguments against said editors, the web poses
308 another limitation: no JavaScript means no
309 editor, and no editor means... (gasp) manually typing in code.</p>
311 <p>Even the most dogmatic purist, however, should recognize that for all
313 There are steps you can take to mitigate the associated drawbacks of
314 these editors.</p>
316 <p>It is often asserted that
319 this is the case with any markup language that allows the smallest
322 A good way to reduce this trouble is to simply eliminate the
323 dialogue boxes that allow users to change colors or fonts (which
324 usually have no legitimate use) and adopt a
326 allowing users to select contextually correct formatting styles
327 for segments of text.</p>
328 </blockquote>
330 <p>Simplicity is also a double-edged sword. The moment any remotely
331 complex markup is needed, these lightweight markup languages fail to
332 produce. Sure you can make '''this text bold''' with Wikitext, but that
335 These languages face the same troubles as regular <acronym
337 whitelist is too restrictive (besides the fact that their table markup
338 is extraordinarily complex).</p>
344 the angled brackets with square brackets and omitting the occasional parameter
345 name? How much more un-original can you get? Somehow, I don't think BBCode
346 was meant to readable. <a
349 <blockquote>
350 BBCode was devised and put to use in order to provide a safer, easier
351 and more limited way of allowing users to format their messages.
352 Previously, many message boards allowed the users to include HTML,
353 which could be used to break/imitate parts of the layout, or run
354 JavaScript. Some implementations of BBCode have suffered problems related
355 to the way they translate the BBCode into HTML, which could negate the
356 security that was intended to be given by BBCode.
357 </blockquote>
361 <blockquote>
362 BBCode came to life when developers where too lazy to parse HTML correctly
363 and decided to invent their own markup language. As with all products of
364 laziness, the result is completely inconsistent, unstandardized, and
365 widely adopted.
366 </blockquote>
368 <p>Well, developers, the whole point of HTML Purifier is that I do the
369 work so you can just execute the ridiculously simple
375 <p>These alternative markup languages have their shiny points, and HTML
376 Purifier is not meant to replace them. However, a major reason for
378 using these languages?</p>
384 neat enough, at least, to make it into PHP as a
386 The premise is simple, the execution effective. Tidy is, in short, a great
389 <p>It is not, however, a filter. I am often surprised when people ask
394 Tidy reads HTML, XHTML and XML files and writes cleaned up markup. For
395 HTML variants, it detects and corrects many common coding errors and
396 strives to produce visually equivalent markup that is both W3C compliant
397 and works on most browsers. A common use of Tidy is to convert plain HTML
398 to XHTML.
399 </blockquote>
408 <p>This is not to say that I haven't seen Tidy be used in this sort of
409 fashion. MediaWiki, for instance, uses Tidy to cleanup the final HTML
410 output before shuttling it off to the browser. The developers, nevertheless,
411 agree that this is only a band-aid solution, and that the real way
412 to fix it is to fix the parser. Tidy's great, but in terms of security,
413 it's not suitable for untrusted sources.</p>
417 <p>I've ordered my analyses according to how bad a library is. The worst
418 is first, and then we move up the spectrum. I will point out the most
419 flagrant problems with the libraries, but note that I will omit more
421 attribute, I really shouldn't reprimand you for letting non-SGML code
422 points through. The ideal solution, however, must do all these things.</p>
424 <p>Note that besides striptags,
425 most of the libraries are moderately effective against the most common XSS
426 attacks. None of them (save Safe HTML Checker) fare very well
427 in the standards-compliance department though.</p>
437 </table>
441 the classic solution for attempting to clean up
444 The fact that it doesn't validate attributes at all means that anyone can
447 <p>While
448 this can be bandaided with a series of regular expressions that strip out
449 on[event] (you're still vulnerable to <acronym
451 quirky browser behavior), striptags() is fundamentally flawed and should not be
452 used.
453 </p>
459 is a souped up version of striptags() with the ability to inspect
460 attributes. (Don't mind the hastily tacked on query escaping function).</p>
473 </table>
475 <p>PHP Input Filter implements an
477 performs very basic checks on whether or not tags and attributes have
478 been defined in the whitelist as well as some smarter <acronym
480 the user to define what they'll permit.</p>
482 <p>With absolutely no checking of well-formedness, it is trivially easy
483 to trick the filter into leaving unclosed tags lying around. While to some
485 basic sanity checks like this must be implemented, otherwise a user
486 can mangle a website's layout.</p>
488 <p>More troubles: Woe to
490 just let <acronym
492 layout not to be badly mutilated. To top things off,
493 the filter doesn't even preserve data properly: attributes have all
494 spaces stripped out of them. Stay away, stay away!</p>
501 filtering library.
502 It should be noted that this is the same library as
504 branding (and a different version number).</p>
517 </table>
519 <p>HTML_Safe's mechanism of action involves parsing
522 validation and filtering as the handlers are called. HTML_Safe does a lot
525 is fundamentally flawed: blacklists.</p>
527 <p>This library maintains arrays of dangerous tags, attributes and
529 has a blacklist of dangerous <acronym
531 intelligently disabled by default in favor of a protocol whitelist.)
532 What this means is that HTML_Safe has no qualms of accepting input
534 the tags in those arrays. Scratch standards-compliance (and that was
535 without even considering proper nesting).</p>
537 <p>For now, HTML_Safe might be safe from
539 In the future, however, one of the infinitely many tags that HTML_Safe lets
540 through might just possibly be given special functionality by browser vendors.
542 solution puts you at a perpetual arms race against crackers who are constantly
543 discovering new and inventive ways to abuse tags and attributes that you
544 didn't blacklist.</p>
549 be the de-facto solution for cleaning
565 </table>
567 <p>To be truthful, I didn't do as comprehensive a code survey for kses
568 as I did for some of the other libraries. Out of
569 all the classes I've reviewed so far, kses was definitely the hardest to
570 understand.</p>
572 <p>kses's modus operandi is splitting up html with a monster regexp
574 suffers from the same problems as Input Filter: no well-formedness
575 checks leading to rampant runaway tags (and no standards-compliance).
576 WordPress, the primary user of kses today, had to implement their
577 own custom tag-balancing code to fix this problem: don't use this
578 library without some equivalent!</p>
580 <p>Its whitelist syntax, however, is the most complex of all these libraries,
581 so I'm going to take some time to argue why this particular implementation
582 is bad. The author of this library was thoughtful enough to provide some
583 basic constraint checks on attributes like maxlen and maxval. Now, barring
584 the fact that there simply aren't enough checks, and the fact that they are
585 all lumped together in one function, we now must wonder whether or not
586 the user will go through the trouble of specifying the maximum length
587 of a title attribute.</p>
589 <p>I have my opinions about inherent human laziness, but perhaps WordPress's
590 default filterset is the most telling example:</p>
592 <pre>
593 $allowedposttags = array (
594 /* formatted and trimmed */
595 'hr' => array (
596 'align' => array (),
597 'noshade' => array (),
598 'size' => array (),
599 'width' => array ()
600 )
601 );
602 </pre>
604 <p>Hmm... do I see a blatant lack of attribute constraints? Conclusion:
605 if the user can get away with not doing work, they will! The biggest
607 the whitelist. The whitelist is just as important as the code that uses
608 the whitelist to filter
615 HTML Checker</a> is (to my knowledge) the first attempt to make a filter
616 that also outputs standards-compliant XHTML. It wasn't even released or
618 search result must have done something right.</p>
631 </table>
633 <p>Indeed, it is quite a well-written piece of code. It demonstrates
634 knowledge of inline versus block elements, thus almost nearly getting
635 nesting correct (the only exception is an unimplemented omitted SGML
638 <p>Unfortunately, part of the reason why it works so well is that it's
639 extremely restrictive. No styling, no tables, very few attributes.
640 Perfectly appropriate for blog comments, but then again, there's always
641 BBCode. This probably means that Safe HTML Checker has a different
642 goal than HTML Purifier.</p>
645 is also quite strict. Accidentally missed a < sign? The parser will
646 complain with the cryptic message:
648 is not well-formed".
649 The solution is not as simple as just switching to a more permissive
650 parser: Safe HTML Checker relies on the fact that the parser will have
651 matched up the tags for them.</p>
666 </table>
677 </table>
679 <p>This is not to say that HTML Purifier doesn't have problems of its own.
680 It's a fairly nascent library (that doesn't mean its buggy though), it's big
681 (while the others usually fit in one file, this one requires a huge
683 features.</a> But even in its current state,
684 HTML Purifier is far better than the other libraries.</p>
688 </div>
689 </body>
690 </html>