Viewing file:  htmlfilter.php (29.34 KB) -rw-r--r--
Select action/file-type:
 (+) |  (+) |  (+) | Code (+) | Session (+) |  (+) | SDB (+) |  (+) |  (+) |  (+) |  (+) |  (+) |
<?php
/**
* htmlfilter.inc
* ---------------
* This set of functions allows you to filter html in order to remove
* any malicious tags from it. Useful in cases when you need to filter
* user input for any cross-site-scripting attempts.
*
* Copyright (C) 2002-2004 by Duke University
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
* 02110-1301 USA
*
* @Author Konstantin Riabitsev <icon@linux.duke.edu>
* @Author Jim Jagielski <jim@jaguNET.com / jimjag@gmail.com>
*/
/**
* This function returns the final tag out of the tag name, an array
* of attributes, and the type of the tag. This function is called by
* tln_sanitize internally.
*
* @param string $tagname the name of the tag.
* @param array $attary the array of attributes and their values
* @param integer $tagtype The type of the tag (see in comments).
* @return string A string with the final tag representation.
*/
function tln_tagprint($tagname, $attary, $tagtype)
{
if ($tagtype == 2) {
$fulltag = '</' . $tagname . '>';
} else {
$fulltag = '<' . $tagname;
if (is_array($attary) && sizeof($attary)) {
$atts = array();
while (list($attname, $attvalue) = each($attary)) {
array_push($atts, "$attname=$attvalue");
}
$fulltag .= ' ' . join(' ', $atts);
}
if ($tagtype == 3) {
$fulltag .= ' /';
}
$fulltag .= '>';
}
return $fulltag;
}
/**
* A small helper function to use with array_walk. Modifies a by-ref
* value and makes it lowercase.
*
* @param string $val a value passed by-ref.
* @return void since it modifies a by-ref value.
*/
function tln_casenormalize(&$val)
{
$val = strtolower($val);
}
/**
* This function skips any whitespace from the current position within
* a string and to the next non-whitespace value.
*
* @param string $body the string
* @param integer $offset the offset within the string where we should start
* looking for the next non-whitespace character.
* @return integer the location within the $body where the next
* non-whitespace char is located.
*/
function tln_skipspace($body, $offset)
{
preg_match('/^(\s*)/s', substr($body, $offset), $matches);
if (sizeof($matches[1])) {
$count = strlen($matches[1]);
$offset += $count;
}
return $offset;
}
/**
* This function looks for the next character within a string. It's
* really just a glorified "strpos", except it catches the failures
* nicely.
*
* @param string $body The string to look for needle in.
* @param integer $offset Start looking from this position.
* @param string $needle The character/string to look for.
* @return integer location of the next occurrence of the needle, or
* strlen($body) if needle wasn't found.
*/
function tln_findnxstr($body, $offset, $needle)
{
$pos = strpos($body, $needle, $offset);
if ($pos === false) {
$pos = strlen($body);
}
return $pos;
}
/**
* This function takes a PCRE-style regexp and tries to match it
* within the string.
*
* @param string $body The string to look for needle in.
* @param integer $offset Start looking from here.
* @param string $reg A PCRE-style regex to match.
* @return array|boolean Returns a false if no matches found, or an array
* with the following members:
* - integer with the location of the match within $body
* - string with whatever content between offset and the match
* - string with whatever it is we matched
*/
function tln_findnxreg($body, $offset, $reg)
{
$matches = array();
$retarr = array();
$preg_rule = '%^(.*?)(' . $reg . ')%s';
preg_match($preg_rule, substr($body, $offset), $matches);
if (!isset($matches[0])) {
$retarr = false;
} else {
$retarr[0] = $offset + strlen($matches[1]);
$retarr[1] = $matches[1];
$retarr[2] = $matches[2];
}
return $retarr;
}
/**
* This function looks for the next tag.
*
* @param string $body String where to look for the next tag.
* @param integer $offset Start looking from here.
* @return array|boolean false if no more tags exist in the body, or
* an array with the following members:
* - string with the name of the tag
* - array with attributes and their values
* - integer with tag type (1, 2, or 3)
* - integer where the tag starts (starting "<")
* - integer where the tag ends (ending ">")
* first three members will be false, if the tag is invalid.
*/
function tln_getnxtag($body, $offset)
{
if ($offset > strlen($body)) {
return false;
}
$lt = tln_findnxstr($body, $offset, '<');
if ($lt == strlen($body)) {
return false;
}
/**
* We are here:
* blah blah <tag attribute="value">
* \---------^
*/
$pos = tln_skipspace($body, $lt + 1);
if ($pos >= strlen($body)) {
return array(false, false, false, $lt, strlen($body));
}
/**
* There are 3 kinds of tags:
* 1. Opening tag, e.g.:
* <a href="blah">
* 2. Closing tag, e.g.:
* </a>
* 3. XHTML-style content-less tag, e.g.:
* <img src="blah"/>
*/
switch (substr($body, $pos, 1)) {
case '/':
$tagtype = 2;
$pos++;
break;
case '!':
/**
* A comment or an SGML declaration.
*/
if (substr($body, $pos + 1, 2) == '--') {
$gt = strpos($body, '-->', $pos);
if ($gt === false) {
$gt = strlen($body);
} else {
$gt += 2;
}
return array(false, false, false, $lt, $gt);
} else {
$gt = tln_findnxstr($body, $pos, '>');
return array(false, false, false, $lt, $gt);
}
break;
default:
/**
* Assume tagtype 1 for now. If it's type 3, we'll switch values
* later.
*/
$tagtype = 1;
break;
}
/**
* Look for next [\W-_], which will indicate the end of the tag name.
*/
$regary = tln_findnxreg($body, $pos, '[^\w\-_]');
if ($regary == false) {
return array(false, false, false, $lt, strlen($body));
}
list($pos, $tagname, $match) = $regary;
$tagname = strtolower($tagname);
/**
* $match can be either of these:
* '>' indicating the end of the tag entirely.
* '\s' indicating the end of the tag name.
* '/' indicating that this is type-3 xhtml tag.
*
* Whatever else we find there indicates an invalid tag.
*/
switch ($match) {
case '/':
/**
* This is an xhtml-style tag with a closing / at the
* end, like so: <img src="blah"/>. Check if it's followed
* by the closing bracket. If not, then this tag is invalid
*/
if (substr($body, $pos, 2) == '/>') {
$pos++;
$tagtype = 3;
} else {
$gt = tln_findnxstr($body, $pos, '>');
$retary = array(false, false, false, $lt, $gt);
return $retary;
}
//intentional fall-through
case '>':
return array($tagname, false, $tagtype, $lt, $pos);
break;
default:
/**
* Check if it's whitespace
*/
if (preg_match('/\s/', $match)) {
} else {
/**
* This is an invalid tag! Look for the next closing ">".
*/
$gt = tln_findnxstr($body, $lt, '>');
return array(false, false, false, $lt, $gt);
}
}
/**
* At this point we're here:
* <tagname attribute='blah'>
* \-------^
*
* At this point we loop in order to find all attributes.
*/
$attary = array();
while ($pos <= strlen($body)) {
$pos = tln_skipspace($body, $pos);
if ($pos == strlen($body)) {
/**
* Non-closed tag.
*/
return array(false, false, false, $lt, $pos);
}
/**
* See if we arrived at a ">" or "/>", which means that we reached
* the end of the tag.
*/
$matches = array();
preg_match('%^(\s*)(>|/>)%s', substr($body, $pos), $matches);
if (isset($matches[0]) && $matches[0]) {
/**
* Yep. So we did.
*/
$pos += strlen($matches[1]);
if ($matches[2] == '/>') {
$tagtype = 3;
$pos++;
}
return array($tagname, $attary, $tagtype, $lt, $pos);
}
/**
* There are several types of attributes, with optional
* [:space:] between members.
* Type 1:
* attrname[:space:]=[:space:]'CDATA'
* Type 2:
* attrname[:space:]=[:space:]"CDATA"
* Type 3:
* attr[:space:]=[:space:]CDATA
* Type 4:
* attrname
*
* We leave types 1 and 2 the same, type 3 we check for
* '"' and convert to """ if needed, then wrap in
* double quotes. Type 4 we convert into:
* attrname="yes".
*/
$regary = tln_findnxreg($body, $pos, '[^\w\-_]');
if ($regary == false) {
/**
* Looks like body ended before the end of tag.
*/
return array(false, false, false, $lt, strlen($body));
}
list($pos, $attname, $match) = $regary;
$attname = strtolower($attname);
/**
* We arrived at the end of attribute name. Several things possible
* here:
* '>' means the end of the tag and this is attribute type 4
* '/' if followed by '>' means the same thing as above
* '\s' means a lot of things -- look what it's followed by.
* anything else means the attribute is invalid.
*/
switch ($match) {
case '/':
/**
* This is an xhtml-style tag with a closing / at the
* end, like so: <img src="blah"/>. Check if it's followed
* by the closing bracket. If not, then this tag is invalid
*/
if (substr($body, $pos, 2) == '/>') {
$pos++;
$tagtype = 3;
} else {
$gt = tln_findnxstr($body, $pos, '>');
$retary = array(false, false, false, $lt, $gt);
return $retary;
}
//intentional fall-through
case '>':
$attary{$attname} = '"yes"';
return array($tagname, $attary, $tagtype, $lt, $pos);
break;
default:
/**
* Skip whitespace and see what we arrive at.
*/
$pos = tln_skipspace($body, $pos);
$char = substr($body, $pos, 1);
/**
* Two things are valid here:
* '=' means this is attribute type 1 2 or 3.
* \w means this was attribute type 4.
* anything else we ignore and re-loop. End of tag and
* invalid stuff will be caught by our checks at the beginning
* of the loop.
*/
if ($char == '=') {
$pos++;
$pos = tln_skipspace($body, $pos);
/**
* Here are 3 possibilities:
* "'" attribute type 1
* '"' attribute type 2
* everything else is the content of tag type 3
*/
$quot = substr($body, $pos, 1);
if ($quot == '\'') {
$regary = tln_findnxreg($body, $pos + 1, '\'');
if ($regary == false) {
return array(false, false, false, $lt, strlen($body));
}
list($pos, $attval, $match) = $regary;
$pos++;
$attary{$attname} = '\'' . $attval . '\'';
} else {
if ($quot == '"') {
$regary = tln_findnxreg($body, $pos + 1, '\"');
if ($regary == false) {
return array(false, false, false, $lt, strlen($body));
}
list($pos, $attval, $match) = $regary;
$pos++;
$attary{$attname} = '"' . $attval . '"';
} else {
/**
* These are hateful. Look for \s, or >.
*/
$regary = tln_findnxreg($body, $pos, '[\s>]');
if ($regary == false) {
return array(false, false, false, $lt, strlen($body));
}
list($pos, $attval, $match) = $regary;
/**
* If it's ">" it will be caught at the top.
*/
$attval = preg_replace('/\"/s', '"', $attval);
$attary{$attname} = '"' . $attval . '"';
}
}
} else {
if (preg_match('|[\w/>]|', $char)) {
/**
* That was attribute type 4.
*/
$attary{$attname} = '"yes"';
} else {
/**
* An illegal character. Find next '>' and return.
*/
$gt = tln_findnxstr($body, $pos, '>');
return array(false, false, false, $lt, $gt);
}
}
}
}
/**
* The fact that we got here indicates that the tag end was never
* found. Return invalid tag indication so it gets stripped.
*/
return array(false, false, false, $lt, strlen($body));
}
/**
* Translates entities into literal values so they can be checked.
*
* @param string $attvalue the by-ref value to check.
* @param string $regex the regular expression to check against.
* @param boolean $hex whether the entites are hexadecimal.
* @return boolean True or False depending on whether there were matches.
*/
function tln_deent(&$attvalue, $regex, $hex = false)
{
preg_match_all($regex, $attvalue, $matches);
if (is_array($matches) && sizeof($matches[0]) > 0) {
$repl = array();
for ($i = 0; $i < sizeof($matches[0]); $i++) {
$numval = $matches[1][$i];
if ($hex) {
$numval = hexdec($numval);
}
$repl{$matches[0][$i]} = chr($numval);
}
$attvalue = strtr($attvalue, $repl);
return true;
} else {
return false;
}
}
/**
* This function checks attribute values for entity-encoded values
* and returns them translated into 8-bit strings so we can run
* checks on them.
*
* @param string $attvalue A string to run entity check against.
* @return Void, modifies a reference value.
*/
function tln_defang(&$attvalue)
{
/**
* Skip this if there aren't ampersands or backslashes.
*/
if (strpos($attvalue, '&') === false
&& strpos($attvalue, '\\') === false
) {
return;
}
do {
$m = false;
$m = $m || tln_deent($attvalue, '/\�*(\d+);*/s');
$m = $m || tln_deent($attvalue, '/\�*((\d|[a-f])+);*/si', true);
$m = $m || tln_deent($attvalue, '/\\\\(\d+)/s', true);
} while ($m == true);
$attvalue = stripslashes($attvalue);
}
/**
* Kill any tabs, newlines, or carriage returns. Our friends the
* makers of the browser with 95% market value decided that it'd
* be funny to make "java[tab]script" be just as good as "javascript".
*
* @param string $attvalue The attribute value before extraneous spaces removed.
* @return Void, modifies a reference value.
*/
function tln_unspace(&$attvalue)
{
if (strcspn($attvalue, "\t\r\n\0 ") != strlen($attvalue)) {
$attvalue = str_replace(
array("\t", "\r", "\n", "\0", " "),
array('', '', '', '', ''),
$attvalue
);
}
}
/**
* This function runs various checks against the attributes.
*
* @param string $tagname String with the name of the tag.
* @param array $attary Array with all tag attributes.
* @param array $rm_attnames See description for tln_sanitize
* @param array $bad_attvals See description for tln_sanitize
* @param array $add_attr_to_tag See description for tln_sanitize
* @return Array with modified attributes.
*/
function tln_fixatts(
$tagname,
$attary,
$rm_attnames,
$bad_attvals,
$add_attr_to_tag
) {
while (list($attname, $attvalue) = each($attary)) {
/**
* See if this attribute should be removed.
*/
foreach ($rm_attnames as $matchtag => $matchattrs) {
if (preg_match($matchtag, $tagname)) {
foreach ($matchattrs as $matchattr) {
if (preg_match($matchattr, $attname)) {
unset($attary{$attname});
continue;
}
}
}
}
/**
* Remove any backslashes, entities, or extraneous whitespace.
*/
tln_defang($attvalue);
tln_unspace($attvalue);
/**
* Now let's run checks on the attvalues.
* I don't expect anyone to comprehend this. If you do,
* get in touch with me so I can drive to where you live and
* shake your hand personally. :)
*/
foreach ($bad_attvals as $matchtag => $matchattrs) {
if (preg_match($matchtag, $tagname)) {
foreach ($matchattrs as $matchattr => $valary) {
if (preg_match($matchattr, $attname)) {
/**
* There are two arrays in valary.
* First is matches.
* Second one is replacements
*/
list($valmatch, $valrepl) = $valary;
$newvalue = preg_replace($valmatch, $valrepl, $attvalue);
if ($newvalue != $attvalue) {
$attary{$attname} = $newvalue;
}
}
}
}
}
}
/**
* See if we need to append any attributes to this tag.
*/
foreach ($add_attr_to_tag as $matchtag => $addattary) {
if (preg_match($matchtag, $tagname)) {
$attary = array_merge($attary, $addattary);
}
}
return $attary;
}
/**
*
* @param string $body The HTML you wish to filter
* @param array $tag_list see description above
* @param array $rm_tags_with_content see description above
* @param array $self_closing_tags see description above
* @param boolean $force_tag_closing see description above
* @param array $rm_attnames see description above
* @param array $bad_attvals see description above
* @param array $add_attr_to_tag see description above
* @return string Sanitized html safe to show on your pages.
*/
function tln_sanitize(
$body,
$tag_list,
$rm_tags_with_content,
$self_closing_tags,
$force_tag_closing,
$rm_attnames,
$bad_attvals,
$add_attr_to_tag
) {
/**
* Normalize rm_tags and rm_tags_with_content.
*/
$rm_tags = array_shift($tag_list);
@array_walk($tag_list, 'tln_casenormalize');
@array_walk($rm_tags_with_content, 'tln_casenormalize');
@array_walk($self_closing_tags, 'tln_casenormalize');
/**
* See if tag_list is of tags to remove or tags to allow.
* false means remove these tags
* true means allow these tags
*/
$curpos = 0;
$open_tags = array();
$trusted = "<!-- begin tln_sanitized html -->\n";
$skip_content = false;
/**
* Take care of netscape's stupid javascript entities like
* &{alert('boo')};
*/
$body = preg_replace('/&(\{.*?\};)/si', '&\\1', $body);
while (($curtag = tln_getnxtag($body, $curpos)) != false) {
list($tagname, $attary, $tagtype, $lt, $gt) = $curtag;
$free_content = substr($body, $curpos, $lt - $curpos);
if ($skip_content == false) {
$trusted .= $free_content;
} else {
}
if ($tagname != false) {
if ($tagtype == 2) {
if ($skip_content == $tagname) {
/**
* Got to the end of tag we needed to remove.
*/
$tagname = false;
$skip_content = false;
} else {
if ($skip_content == false) {
if (isset($open_tags{$tagname}) &&
$open_tags{$tagname} > 0
) {
$open_tags{$tagname}--;
} else {
$tagname = false;
}
} else {
}
}
} else {
/**
* $rm_tags_with_content
*/
if ($skip_content == false) {
/**
* See if this is a self-closing type and change
* tagtype appropriately.
*/
if ($tagtype == 1
&& in_array($tagname, $self_closing_tags)
) {
$tagtype = 3;
}
/**
* See if we should skip this tag and any content
* inside it.
*/
if ($tagtype == 1
&& in_array($tagname, $rm_tags_with_content)
) {
$skip_content = $tagname;
} else {
if (($rm_tags == false
&& in_array($tagname, $tag_list)) ||
($rm_tags == true
&& !in_array($tagname, $tag_list))
) {
$tagname = false;
} else {
if ($tagtype == 1) {
if (isset($open_tags{$tagname})) {
$open_tags{$tagname}++;
} else {
$open_tags{$tagname} = 1;
}
}
/**
* This is where we run other checks.
*/
if (is_array($attary) && sizeof($attary) > 0) {
$attary = tln_fixatts(
$tagname,
$attary,
$rm_attnames,
$bad_attvals,
$add_attr_to_tag
);
}
}
}
} else {
}
}
if ($tagname != false && $skip_content == false) {
$trusted .= tln_tagprint($tagname, $attary, $tagtype);
}
} else {
}
$curpos = $gt + 1;
}
$trusted .= substr($body, $curpos, strlen($body) - $curpos);
if ($force_tag_closing == true) {
foreach ($open_tags as $tagname => $opentimes) {
while ($opentimes > 0) {
$trusted .= '</' . $tagname . '>';
$opentimes--;
}
}
$trusted .= "\n";
}
$trusted .= "<!-- end tln_sanitized html -->\n";
return $trusted;
}
//
// Use the nifty htmlfilter library
//
function HTMLFilter($body, $trans_image_path, $block_external_images = false)
{
$tag_list = array(
false,
"object",
"meta",
"html",
"head",
"base",
"link",
"frame",
"iframe",
"plaintext",
"marquee"
);
$rm_tags_with_content = array(
"script",
"applet",
"embed",
"title",
"frameset",
"xmp",
"xml"
);
$self_closing_tags = array(
"img",
"br",
"hr",
"input",
"outbind"
);
$force_tag_closing = true;
$rm_attnames = array(
"/.*/" =>
array(
// "/target/i",
"/^on.*/i",
"/^dynsrc/i",
"/^data.*/i",
"/^lowsrc.*/i"
)
);
$bad_attvals = array(
"/.*/" =>
array(
"/^src|background/i" =>
array(
array(
'/^([\'"])\s*\S+script\s*:.*([\'"])/si',
'/^([\'"])\s*mocha\s*:*.*([\'"])/si',
'/^([\'"])\s*about\s*:.*([\'"])/si'
),
array(
"\\1$trans_image_path\\2",
"\\1$trans_image_path\\2",
"\\1$trans_image_path\\2",
"\\1$trans_image_path\\2"
)
),
"/^href|action/i" =>
array(
array(
'/^([\'"])\s*\S+script\s*:.*([\'"])/si',
'/^([\'"])\s*mocha\s*:*.*([\'"])/si',
'/^([\'"])\s*about\s*:.*([\'"])/si'
),
array(
"\\1#\\1",
"\\1#\\1",
"\\1#\\1",
"\\1#\\1"
)
),
"/^style/i" =>
array(
array(
"/expression/i",
"/binding/i",
"/behaviou*r/i",
"/include-source/i",
'/position\s*:\s*absolute/i',
'/url\s*\(\s*([\'"])\s*\S+script\s*:.*([\'"])\s*\)/si',
'/url\s*\(\s*([\'"])\s*mocha\s*:.*([\'"])\s*\)/si',
'/url\s*\(\s*([\'"])\s*about\s*:.*([\'"])\s*\)/si',
'/(.*)\s*:\s*url\s*\(\s*([\'"]*)\s*\S+script\s*:.*([\'"]*)\s*\)/si'
),
array(
"idiocy",
"idiocy",
"idiocy",
"idiocy",
"",
"url(\\1#\\1)",
"url(\\1#\\1)",
"url(\\1#\\1)",
"url(\\1#\\1)",
"url(\\1#\\1)",
"\\1:url(\\2#\\3)"
)
)
)
);
if ($block_external_images) {
array_push(
$bad_attvals{'/.*/'}{'/^src|background/i'}[0],
'/^([\'\"])\s*https*:.*([\'\"])/si'
);
array_push(
$bad_attvals{'/.*/'}{'/^src|background/i'}[1],
"\\1$trans_image_path\\1"
);
array_push(
$bad_attvals{'/.*/'}{'/^style/i'}[0],
'/url\(([\'\"])\s*https*:.*([\'\"])\)/si'
);
array_push(
$bad_attvals{'/.*/'}{'/^style/i'}[1],
"url(\\1$trans_image_path\\1)"
);
}
$add_attr_to_tag = array(
"/^a$/i" =>
array('target' => '"_blank"')
);
$trusted = tln_sanitize(
$body,
$tag_list,
$rm_tags_with_content,
$self_closing_tags,
$force_tag_closing,
$rm_attnames,
$bad_attvals,
$add_attr_to_tag
);
return $trusted;
}
|