Difference between revisions of "SpamFerret.php"

from HTYP, the free directory anyone can edit if they can prove to me that they're not a spambot
Jump to navigation Jump to search
(7/14 version: uses LibMgr, adds attempts.Diff field)
(→‎Code: fixed lots of warnings; FigureDIff() function instead of xdiff (which isn't standard))
Line 102: Line 102:
 
  * Hook function for $wgFilterCallback
 
  * Hook function for $wgFilterCallback
 
  */
 
  */
−
function wfSpamFerretFilter( &$title, $text, $section, &$hookErr, $editSummary ) {
+
//function wfSpamFerretFilter( &$title, $text, $section, &$hookErr, $editSummary ) {
 +
function wfSpamFerretFilter( &$title, $text, $section ) {
 
     global $wgOut;
 
     global $wgOut;
  
 
     $spamObj = GetSpamFerret();
 
     $spamObj = GetSpamFerret();
 
     $wgOut->addWikiText( "Intercepted by SpamFerretFilter" );
 
     $wgOut->addWikiText( "Intercepted by SpamFerretFilter" );
 +
//    $ret = $spamObj->filter( $title, $text, '', $editSummary, $editPage );
 
     return $spamObj->filter( $title, $text, $section );
 
     return $spamObj->filter( $title, $text, $section );
 
}
 
}
Line 124: Line 126:
 
     $spamObj = GetSpamFerret();
 
     $spamObj = GetSpamFerret();
 
     $title = $editPage->mArticle->getTitle();
 
     $title = $editPage->mArticle->getTitle();
−
    $ret = $spamObj->filter( $title, $text, '', $editSummary, $editPage );
+
//    $ret = $spamObj->filter( $title, $text, '', $editSummary, $editPage );
 +
    $ret = $spamObj->filter( $title, $text, '' );
 
     if ( $ret !== false ) $editPage->spamPage( $ret );
 
     if ( $ret !== false ) $editPage->spamPage( $ret );
  
Line 158: Line 161:
 
}
 
}
  
−
function filter( &$title, $text, $section, $editSummary, $editPage ) {
+
// function filter( &$title, $text, $section, $editSummary, $editPage ) {
 +
function filter( &$title, $text, $section ) {
 
global $wgArticle, $wgDBname, $wgMemc, $messageMemc, $wgVersion, $wgOut;
 
global $wgArticle, $wgDBname, $wgMemc, $messageMemc, $wgVersion, $wgOut;
 
global $wgTitle, $wgServer;
 
global $wgTitle, $wgServer;
Line 232: Line 236:
 
if ($objArticleCurr->exists()) {
 
if ($objArticleCurr->exists()) {
 
$txtCurr = $objArticleCurr->getContent();
 
$txtCurr = $objArticleCurr->getContent();
−
$this->txtDiff = xdiff_string_diff($text,$txtCurr);
+
$this->txtDiff = FigureDiff($txtCurr,$text);
 
$lenIns = strpos($text,$txtCurr);
 
$lenIns = strpos($text,$txtCurr);
 
if ($lenIns !== false) {
 
if ($lenIns !== false) {
Line 241: Line 245:
 
$strIns = $text;
 
$strIns = $text;
 
$txtCurr = NULL;
 
$txtCurr = NULL;
−
$this->txtDiff = xdiff_string_diff($text,'');
+
$this->txtDiff = FigureDiff('',$text);
 
}
 
}
 
// ** OFFENSE: Spam turds (short bits of nonsense inserted at the beginning of an article):
 
// ** OFFENSE: Spam turds (short bits of nonsense inserted at the beginning of an article):
Line 441: Line 445:
 
}
 
}
 
}
 
}
 +
}
 +
function FigureDiff($iTextOld, $iTextNew) {
 +
# Make temporary files
 +
$td = wfTempDir();
 +
$fhOld = fopen( $fnOld = tempnam( $td, 'SpamFerret-old-' ), 'w' );
 +
$fhNew = fopen( $fnNew = tempnam( $td, 'SpamFerret-new-' ), 'w' );
 +
 +
fwrite( $fhOld, $iTextOld ); fclose( $fhOld );
 +
fwrite( $fhNew, $iTextNew ); fclose( $fhNew );
 +
 +
$ksSpamFerretDiffOptions = '-i -E --suppress-common-lines'; // this setting applies to all uses of a given filter DB
 +
// can change for different DBs
 +
$cmd = 'diff '.$ksSpamFerretDiffOptions.wfEscapeShellArg( $fnOld, $fnNew );
 +
$handle = popen( $cmd, 'r' );
 +
$result = '';
 +
do {
 +
$data = fread( $handle, 8192 );
 +
if ( strlen( $data ) == 0 ) {
 +
break;
 +
}
 +
$result .= $data;
 +
} while ( true );
 +
pclose( $handle );
 +
unlink( $mytextName ); unlink( $oldtextName ); unlink( $yourtextName );
 +
return $result;
 
}
 
}
  
 
} // end of 'MEDIAWIKI' check
 
} // end of 'MEDIAWIKI' check

Revision as of 17:06, 14 July 2009

Navigation

{{#lst:SpamFerret|navbar}}: SpamFerret.php

Code

<php><?php /* 

HISTORY:
 2007-08-19 (Wzl) fixed line 155 call to clsDatabase::Query() (method deprecated and removed)
 2007-09-30 (Wzl) fixing regex processing
 2007-10-11 (Wzl) logging ampersandbot attempts; spam throttle (automatic temporary blacklist)
 2007-10-13 (Wzl) fixed some issues which were preventing throttling from working - mainly changes to SQL
 2007-10-15 (Wzl) "Code" wasn't being recorded. Decided that normal filtering should use code '-' so NULL means something is wrong.
 2007-10-28 (Wzl) Events with THR and AMP codes *still* weren't being recorded because method call was improperly formatted.
 2007-12-23 (Wzl) Emails wikimaster if eregi() returns an error (due to improperly formatted regex)
 2007-12-26 (Wzl) Spam turd rejection / logging
 2007-12-27 (Wzl) Fixed spam turd detection to work for new pages too (will probably need refinement)
 2007-12-27 (Wzl) Ooops. Replaced missing if-block from regex results when inserted text is found.
 2008-08-29 (Wzl) Added permanent IP blocking
 2008-09-04 (Wzl) Added (optional) logging of successful edits
 2008-09-19 (Wzl) Actually *set* the "didEdit" flag for successful edits <facepalms>
 2008-10-21 (Wzl) Fixed minor syntax error in "defines"
 2009-02-25 (Wzl) $objArticleCurr->loadLastEdit() now causes error in MW 1.14 (was it necessary before?)
 2009-03-10 (Wzl) "require" -> "require_once" so other extensions can use data.php without conflict
   also optional $kfpWzlLibs so data.php can be somewhere not on the path
 2009-03-18 (Wzl) Got rid of shared.php requirement; now using kfpLib to locate data.php
 2009-03-26 (Wzl) Rewrote data library calls to use new classes (no longer using deprecated/removed classes)
   Also modified to use newer function hooks
 2009-04-24 (Wzl) fixed "strict" bug referencing unset $txtCurr when creating new page
 2009-07-05 (Wzl) Using LibMgr
 2009-07-14 (Wzl) Added attempts.Diff field
TO DO:
 * Log matching text for regex filters
 * Throttled save attempts should check for spam, just for data-gathering purposes.
   Possibly non-spam from a throttled IP should not update the "WhenLast" timestamp. Maybe this should be a LocalSettings option?
 * Figure out how to display a different error message than "the following text is what triggered our filter:"
OPTIONAL SETTINGS:
 kfpLib - path to data.php folder (no final slash)
 kfsLib_Data - filespec of data.php
  • /
  1. Loader for spam blacklist feature
  2. Include this from LocalSettings.php

if ( defined( 'MEDIAWIKI' ) ) {

$wgExtensionCredits['other'][] = array( 'name' => 'SpamFerret', 'author' => 'Woozle Staddon', 'url' => 'http://htyp.org/SpamFerret', 'version' => '2009-07-14', 'description' => 'database-driven wikispam blocker', ); if (!defined('kfsLib_Data')) { 

   if (defined('kfpLib')) {

define('kfsLib_Data', kfpLib.'/data.php'); 

   } else {

define('kfsLib_Data','data.php'); // assume it's on the path 

   }

} if (!defined('LIBMGR')) { 

   require('libmgr.php');

} clsLibMgr::Add('data', kfsLib_Data); clsLibMgr::Load('data');

/* ============== 

SET UP CALLBACKS
  • /

global $wgFilterCallback, $wgPreSpamFilterCallback;

$wgPreSpamFilterCallback = false; if ( defined( 'MW_SUPPORTS_EDITFILTERMERGED' ) ) { 

   $wgHooks['EditFilterMerged'][] = 'wfSpamFerretMerged';

} else { 

   if ( $wgFilterCallback ) {
       $wgPreSpamFilterCallback = $wgFilterCallback;
   }
   $wgFilterCallback = 'wfSpamFerretFilter';

}

/* $wgHooks['EditFilter'][] = 'wfSpamFerretValidate'; $wgHooks['ArticleSaveComplete'][] = 'wfSpamFerretArticleSave'; $wgHooks['APIEditBeforeSave'][] = 'wfSpamFerretAPIEditBeforeSave';

  • /

/* ================ 

SET GLOBAL OBJECTS
  • /

function GetSpamFerret() { 

   static $objFerret;

   if (!isset($objFerret)) {

$objFerret = new SpamFerret(); 

   }
   return $objFerret;

} /* ================ 

CALLBACK FUNCTIONS
  • /

/** 

* Hook function for $wgFilterCallback
*/

//function wfSpamFerretFilter( &$title, $text, $section, &$hookErr, $editSummary ) { function wfSpamFerretFilter( &$title, $text, $section ) { 

   global $wgOut;

   $spamObj = GetSpamFerret();
   $wgOut->addWikiText( "Intercepted by SpamFerretFilter" );

// $ret = $spamObj->filter( $title, $text, , $editSummary, $editPage ); 

   return $spamObj->filter( $title, $text, $section );

}

/** 

* Hook function for EditFilterMerged, replaces wfSpamBlacklistFilter
*/

function wfSpamFerretMerged( &$editPage, $text, &$hookErr, $editSummary ) { 

   global $wgTitle,$wgOut;

   if( is_null( $wgTitle ) ) {
       # API mode
       # wfSpamBlacklistFilterAPIEditBeforeSave already checked the blacklist
       return true;
   }

   $spamObj = GetSpamFerret();
   $title = $editPage->mArticle->getTitle();

// $ret = $spamObj->filter( $title, $text, , $editSummary, $editPage ); 

   $ret = $spamObj->filter( $title, $text,  );
   if ( $ret !== false ) $editPage->spamPage( $ret );

// additional text can be added here: // $wgOut->addWikiText( "Intercepted by SpamFerretMerged" ); 

   // Return convention for hooks is the inverse of $wgFilterCallback
   return ( $ret === false );

} /** 

* Hook function for APIEditBeforeSave
*/

function wfSpamFerretAPIEditBeforeSave( &$editPage, $text, &$resultArr ) { }

class SpamFerret { var $previousFilter = false; // internal data var $dbSpam; var $objDataClients; var $strIPAddr; var $idPattern; var $idClient; var $doClearThrottle;

// function SpamFerret() { // }

function Setting($iName) { global $wgSpamFerretSettings;

return $wgSpamFerretSettings[$iName]; }

// function filter( &$title, $text, $section, $editSummary, $editPage ) { function filter( &$title, $text, $section ) { global $wgArticle, $wgDBname, $wgMemc, $messageMemc, $wgVersion, $wgOut; global $wgTitle, $wgServer; global $wgEmergencyContact; global $debug; global $errNum, $errStr; // debugging: global $sql;

$fname = 'wfSpamFerretFilter'; wfProfileIn( $fname );

# Call the rest of the hook chain first if ( $this->previousFilter ) { $f = $this->previousFilter; if ( $f( $title, $text, $section ) ) { wfProfileOut( $fname ); return true; } } // initialize variables $retVal = FALSE; // default = assume edit is ok $msgEmail = FALSE;

// get the IP address of the http client making the edit attempt: $this->strIPAddr = wfGetIP(); // Open the database $this->dbSpam = new clsDatabase($this->Setting('dbspec')); $this->dbSpam->Open(); // open clients table (extended Throttle version) for reference: //return TRUE; $objTblClients = new clsTable($this->dbSpam,'ClientThrottle','ID'); // Look up to see if this IP is known; it may already be throttled: $this->objDataClients = $objTblClients->GetData('Address="'.$this->strIPAddr.'"'); if (is_object($this->objDataClients)) { if ($this->objDataClients->hasRows()) { $isClientKnown = true; } } if ($isClientKnown) { $this->idClient = $this->objDataClients->ID; $doBlock = $this->objDataClients->doBlock; if ($doBlock) { $strThrType = 'BLK'; } else { $intRetries = $this->objDataClients->Retries; $intThrottle = $this->Setting('throttle_retries'); $doBlock = $intRetries > $intThrottle; $strThrType = 'THR-'.$intRetries; } if ($doBlock) { // retry limit exceeded; check timeout limit if ($this->objDataClients->ThrottleTime < $this->Setting('throttle_timeout')) { EditPage::spamPage('Too many spam attempts from your IP address ('.$this->strIPAddr.'). Please come back later.'); $this->RecordAttempt($strThrType); // record post attempt by throttled client $retVal = true; // client has exceeded spam limit; impose throttle } else { $this->doClearThrottle = true; } } }

if (!$retVal) { /* 

At this point, there's apparently no reason to block the client just for being who they are,

so now check for common non-listable offenses. These involve comparing the new contents with the original, so first we get the original (current) article contents plus some information about what has changed: 

* $strIns = whatever has been inserted at the start of the article (or contents of new article)
  • /

$objArticleCurr = new Article($title); //$objArticleCurr->loadLastEdit(); // function is now protected if ($objArticleCurr->exists()) { $txtCurr = $objArticleCurr->getContent(); $this->txtDiff = FigureDiff($txtCurr,$text); $lenIns = strpos($text,$txtCurr); if ($lenIns !== false) { $strIns = substr($text,0,$lenIns); } } else { $lenIns = strlen($text); $strIns = $text; $txtCurr = NULL; $this->txtDiff = FigureDiff(,$text); } // ** OFFENSE: Spam turds (short bits of nonsense inserted at the beginning of an article): if ($strIns != ) { // new page is old page with something inserted at the beginning // get the inserted text $isMatch = preg_match('/^[a-z0-9]+ ?$/',$strIns); // another way to do it, incomplete: // $lenMatch = strspn($strIns, '1234567890abcdefghijklmnopqrstuvwxyz'); // if (strlen($strIns) - $lenMatch < 2) { // } if ($isMatch) { $this->RecordAttempt('TRD',$strIns); // record spam attempt (AMP = ampersandbot) // The string sent to spamPage() will be shown after "The following text is what triggered our spam filter:" EditPage::spamPage( '"'.$strIns.'" (spam turd).' ); $retVal = true; } } if (!is_null($txtCurr)) { // ** OFFENSE: Ampersandbot: $lenNew = strlen($text); $posMatch = strpos($txtCurr, $text); if ($posMatch===0) { // new string starts the same as old string; is it a truncated subset? if ($lenNew < strlen($txtCurr)) { // new string is a truncation of old string // ideally, we would just check to see if the missing character is an ampersand - // ...but unfortunately, something is quasi-randomly mutating the strings in a way which // leaves the exact position of the "missing character" in some doubt. So what we do is this: // 1. Find the position of the first ampersand in OLD TEXT: $posAmp = strpos($txtCurr,'&'); // 2. Compare this position with the length of NEW TEXT: $posDiff = abs(strlen($text)-$posAmp); // 3. If the difference is less than some limit, then presume Ampersandbot activity: if ($posDiff < 3) { // TO DO: log $posDiff for later analysis // AMPERSANDBOT DETECTED; refuse to save the edit $this->RecordAttempt('AMP'); // record spam attempt (AMP = ampersandbot) // The string returned will be shown after "The following text is what triggered our spam filter:" $retVal = 'The removal of everything after the first ampersand. You are an ampersandbot; please go away.'; // LATER: it would be nice to display the error in a different format from the usual spam page } } } } } if ($debug) { EditPage::spamPage('DEBUG: '.$debug); $retVal = true; }

if (!$retVal) { // set_error_handler ('ErrorHandler',E_WARNING); $objTblPatterns = new clsTable($this->dbSpam,'patterns','ID'); $objDataPatterns = $objTblPatterns->GetData('isActive');

$strTextCk = strtolower($text); $isMatch = FALSE; while($objDataPatterns->NextRow() && !$isMatch) { $strPattern = $objDataPatterns->Pattern; $isRegex = $objDataPatterns->isRegex; $this->idPattern = $objDataPatterns->ID; if ($isRegex) { $strPattCk = $strPattern; // Attempt at using Perl-compatible regex, but it doesn't seem to work (or maybe just too many patterns have issues): // $strPattCk_pcre = str_replace('/','\/',$strPattCk); // $isMatch = preg_match('/'.$strPattCk_pcre.'/i',$strTextCk,$matches); $isMatch = eregi($strPattCk, $strTextCk, $matches); if ($errNum) { $msgEmail .= 'Filter #'.$this->idPattern.' generated error #'.$errNum.': '.$errStr."\n"; // echo '
ERROR #'.$errNum.': '.$errStr; $errNum = 0; }

if ($isMatch) { $strMatch = $matches[0]; //echo ' match: '.$strMatch; } } else { //echo ' non-regex'; $strMatch = stristr ($strTextCk,$strPattern); $isMatch = ($strMatch != ); //echo ' match: '.$strMatch; } // $debug .= 'ROW: '.DumpArray($objDataPatterns->Row); }

if ( $strMatch != ) { // spam cue found; display the matching text and don't allow the edit to be saved: wfDebug( "Match!\n" );

// The string sent to spamPage() will be shown after "The following text is what triggered our spam filter:" $retVal = '(pattern #'.$this->idPattern.') '.$strMatch; // Log the spam attempt: // $sql = 'SELECT * FROM clients WHERE Address="'.$this->strIPAddr.'"'; // $this->objDataClients = $this->dbSpam->Query($sql); // update or create client record: $this->RecordAttempt('-'); } else { // no spam cues found; allow the edit to be saved, if nothing else has tripped the filter if ($this->Setting('log_ok_edits')) { $this->RecordOkEdit(); } /* EditPage::spamPage( 'DEBUGGING: '.$debug ); $retVal = true; /*/ /**/ } $objDataPatterns->NextRow(); }

wfProfileOut( $fname ); if ($msgEmail) { mail ($wgEmergencyContact,'spamferret filter error',$msgEmail); // die(); } //$wgOut->addHTML($out); return $retVal; /**/ } public function RecordAttempt($iCode,$iMatch=NULL) { global $wgTitle, $wgServer;

if ($this->idClient != 0) { if ($this->doClearThrottle) { $strRetries = '0'; } else { $strRetries = 'Retries+1'; } $sql = 'UPDATE clients SET WhenLast=NOW(),Count=Count+1, Retries='.$strRetries.' WHERE Address="'.$this->strIPAddr.'"'; $this->dbSpam->Exec($sql); } else { $sql = 'INSERT INTO clients (Address,WhenFirst,Count,Retries) VALUES("'.$this->strIPAddr.'",NOW(),1,0)'; $this->dbSpam->Exec($sql); $this->idClient = $this->dbSpam->NewID(); } $sqlCode = '"'.$iCode.'"'; $sqlURL = '"'.$this->dbSpam->SafeParam($wgTitle->getFullURL()).'"'; $sqlSrvr = '"'.$this->dbSpam->SafeParam($wgServer).'"'; $sqlPage = '"'.$this->dbSpam->SafeParam($wgTitle->getPrefixedText()).'"'; $sqlDiff = '"'.$this->dbSpam->SafeParam($this->txtDiff).'"'; $sqlPattern = $this->idPattern; if (is_null($sqlPattern)) { $sqlPattern = 'NULL'; } else { $isPattern = true; } $sqlMatch = is_null($iMatch)?'NULL':'"'.$iMatch.'"'; $sql = 'INSERT INTO attempts (`When`,ID_Pattern,ID_Client,IDS_Session,PageServer,PageName,Code,MatchText,didAllow,Diff) VALUES' .'(NOW(),' .$sqlPattern.',' .$this->idClient.',' .SQL_Value(session_id()).',' .$sqlSrvr.',' .$sqlPage.',' .$sqlCode.',' .$sqlMatch.',' .'FALSE,' .$sqlDiff.')'; $this->dbSpam->Exec($sql); if ($isPattern) { $sql = 'UPDATE patterns SET WhenTried=NOW(), Count=Count+1 WHERE ID='.$this->idPattern; $this->dbSpam->Exec($sql); } } public function RecordOkEdit() { global $wgTitle, $wgServer;

$sqlPage = '"'.$this->dbSpam->SafeParam($wgTitle->getPrefixedText()).'"'; $sqlSrvr = '"'.$this->dbSpam->SafeParam($wgServer).'"'; $sqlDiff = '"'.$this->dbSpam->SafeParam($this->txtDiff).'"'; $sql = 'INSERT INTO attempts (`When`,ID_Pattern,ID_Client,IDS_Session,PageServer,PageName,Code,MatchText,didAllow,Diff) VALUES ' .'(NOW(),NULL,' .$this->idClient.',' .SQL_Value(session_id()).',' .$sqlSrvr.',' .$sqlPage.',NULL,NULL,TRUE,' .$sqlDiff.')'; $this->dbSpam->Exec($sql); } }

function ErrorHandler ($errno ,$errstr) { global $errNum, $errStr;

$errNum = $errno; $errStr = $errstr; }

function SQL_Value($iVar) { if (is_null($iVar)) { return 'NULL'; } else { if (is_numeric($iVar)) { return $iVar; } else { return '"'.$iVar.'"'; } } } function FigureDiff($iTextOld, $iTextNew) { # Make temporary files $td = wfTempDir(); $fhOld = fopen( $fnOld = tempnam( $td, 'SpamFerret-old-' ), 'w' ); $fhNew = fopen( $fnNew = tempnam( $td, 'SpamFerret-new-' ), 'w' );

fwrite( $fhOld, $iTextOld ); fclose( $fhOld ); fwrite( $fhNew, $iTextNew ); fclose( $fhNew );

$ksSpamFerretDiffOptions = '-i -E --suppress-common-lines'; // this setting applies to all uses of a given filter DB // can change for different DBs $cmd = 'diff '.$ksSpamFerretDiffOptions.wfEscapeShellArg( $fnOld, $fnNew ); $handle = popen( $cmd, 'r' ); $result = ; do { $data = fread( $handle, 8192 ); if ( strlen( $data ) == 0 ) { break; } $result .= $data; } while ( true ); pclose( $handle ); unlink( $mytextName ); unlink( $oldtextName ); unlink( $yourtextName ); return $result; }

} // end of 'MEDIAWIKI' check

Retrieved from "https://htyp.org/mw/index.php?title=SpamFerret.php&oldid=12017"