spam/spam.php

   1 <?php
   2 // $Id: spam.php,v 1.60 2006/12/04 13:34:17 henoheno Exp $
   3 // Copyright (C) 2006 PukiWiki Developers Team
   4 // License: GPL v2 or (at your option) any later version
   5
   6 // Functions for Concept-work of spam-uri metrics
   7
   8 if (! defined('SPAM_INI_FILE')) define('SPAM_INI_FILE', 'spam.ini.php');
   9
  10 // ---------------------
  11 // URI pickup
  12
  13 // Return an array of URIs in the $string
  14 // [OK] http://nasty.example.org#nasty_string
  15 // [OK] http://nasty.example.org:80/foo/xxx#nasty_string/bar
  16 // [OK] ftp://nasty.example.org:80/dfsdfs
  17 // [OK] ftp://cnn.example.com&story=breaking_news@10.0.0.1/top_story.htm (from RFC3986)
  18 function uri_pickup($string = '', $normalize = TRUE,
  19         $preserve_rawuri = FALSE, $preserve_chunk = TRUE)
  20 {
  21         // Not available for: IDN(ignored)
  22         $array = array();
  23         preg_match_all(
  24                 // scheme://userinfo@host:port/path/or/pathinfo/maybefile.and?query=string#fragment
  25                 // Refer RFC3986 (Regex below is not strict)
  26                 '#(\b[a-z][a-z0-9.+-]{1,8})://' .       // 1: Scheme
  27                 '(?:' .
  28                         '([^\s<>"\'\[\]/\#?@]*)' .              // 2: Userinfo (Username)
  29                 '@)?' .
  30                 '(' .
  31                         // 3: Host
  32                         '\[[0-9a-f:.]+\]' . '|' .                               // IPv6([colon-hex and dot]): RFC2732
  33                         '(?:[0-9]{1-3}\.){3}[0-9]{1-3}' . '|' . // IPv4(dot-decimal): 001.22.3.44
  34                         '[^\s<>"\'\[\]:/\#?]+' .                                // FQDN: foo.example.org
  35                 ')' .
  36                 '(?::([0-9]*))?' .                                      // 4: Port
  37                 '((?:/+[^\s<>"\'\[\]/\#]+)*/+)?' .      // 5: Directory path or path-info
  38                 '([^\s<>"\'\[\]\#?]+)?' .                       // 6: File?
  39                 '(?:\?([^\s<>"\'\[\]\#]+))?' .          // 7: Query string
  40                 '(?:\#([a-z0-9._~%!$&\'()*+,;=:@-]*))?' .       // 8: Fragment
  41                 '#i',
  42                  $string, $array, PREG_SET_ORDER | PREG_OFFSET_CAPTURE
  43         );
  44         //var_dump(recursive_map('htmlspecialchars', $array));
  45
  46         // Shrink $array
  47         static $parts = array(
  48                 1 => 'scheme', 2 => 'userinfo', 3 => 'host', 4 => 'port',
  49                 5 => 'path', 6 => 'file', 7 => 'query', 8 => 'fragment'
  50         );
  51         $default = array('');
  52         foreach(array_keys($array) as $uri) {
  53                 $_uri = & $array[$uri];
  54                 array_rename_keys($_uri, $parts, TRUE, $default);
  55
  56                 $offset = $_uri['scheme'][1]; // Scheme's offset
  57                 foreach(array_keys($_uri) as $part) {
  58                         // Remove offsets for each part
  59                         $_uri[$part] = & $_uri[$part][0];
  60                 }
  61
  62                 if ($normalize) {
  63                         $_uri['scheme'] = scheme_normalize($_uri['scheme']);
  64                         if ($_uri['scheme'] === '') {
  65                                 unset ($array[$uri]);
  66                                 continue;
  67                         }
  68                         $_uri['host']  = strtolower($_uri['host']);
  69                         $_uri['port']  = port_normalize($_uri['port'], $_uri['scheme'], FALSE);
  70                         $_uri['path']  = path_normalize($_uri['path']);
  71                         if ($preserve_rawuri) $_uri['rawuri'] = & $_uri[0];
  72
  73                         // DEBUG
  74                         //$_uri['uri'] = uri_array_implode($_uri);
  75                 } else {
  76                         $_uri['uri'] = & $_uri[0]; // Raw
  77                 }
  78                 unset($_uri[0]); // Matched string itself
  79                 if (! $preserve_chunk) {
  80                         unset(
  81                                 $_uri['scheme'],
  82                                 $_uri['userinfo'],
  83                                 $_uri['host'],
  84                                 $_uri['port'],
  85                                 $_uri['path'],
  86                                 $_uri['file'],
  87                                 $_uri['query'],
  88                                 $_uri['fragment']
  89                         );
  90                 }
  91                 $_uri['area']['offset'] = $offset;
  92         }
  93
  94         return $array;
  95 }
  96
  97 // Destructive normalize of URI array
  98 // NOTE: Give me the uri_pickup() result with chunks
  99 function uri_array_normalize(& $pickups, $preserve = TRUE)
 100 {
 101         if (! is_array($pickups)) return $pickups;
 102
 103         foreach (array_keys($pickups) as $key) {
 104                 $_key = & $pickups[$key];
 105                 $_key['path']     = isset($_key['path']) ? strtolower($_key['path']) : '';
 106                 $_key['file']     = isset($_key['file']) ? file_normalize($_key['file']) : '';
 107                 $_key['query']    = isset($_key['query']) ? query_normalize(strtolower($_key['query']), TRUE) : '';
 108                 $_key['fragment'] = (isset($_key['fragment']) && $preserve) ?
 109                         strtolower($_key['fragment']) : ''; // Just ignore
 110         }
 111
 112         return $pickups;
 113 }
 114
 115 // An URI array => An URI (See uri_pickup())
 116 function uri_array_implode($uri = array())
 117 {
 118         if (empty($uri) || ! is_array($uri)) return NULL;
 119
 120         $tmp = array();
 121         if (isset($uri['scheme']) && $uri['scheme'] !== '') {
 122                 $tmp[] = & $uri['scheme'];
 123                 $tmp[] = '://';
 124         }
 125         if (isset($uri['userinfo']) && $uri['userinfo'] !== '') {
 126                 $tmp[] = & $uri['userinfo'];
 127                 $tmp[] = '@';
 128         }
 129         if (isset($uri['host']) && $uri['host'] !== '') {
 130                 $tmp[] = & $uri['host'];
 131         }
 132         if (isset($uri['port']) && $uri['port'] !== '') {
 133                 $tmp[] = ':';
 134                 $tmp[] = & $uri['port'];
 135         }
 136         if (isset($uri['path']) && $uri['path'] !== '') {
 137                 $tmp[] = & $uri['path'];
 138         }
 139         if (isset($uri['file']) && $uri['file'] !== '') {
 140                 $tmp[] = & $uri['file'];
 141         }
 142         if (isset($uri['query']) && $uri['query'] !== '') {
 143                 $tmp[] = '?';
 144                 $tmp[] = & $uri['query'];
 145         }
 146         if (isset($uri['fragment']) && $uri['fragment'] !== '') {
 147                 $tmp[] = '#';
 148                 $tmp[] = & $uri['fragment'];
 149         }
 150
 151         return implode('', $tmp);
 152 }
 153
 154 // $array['something'] => $array['wanted']
 155 function array_rename_keys(& $array, $keys = array('from' => 'to'), $force = FALSE, $default = '')
 156 {
 157         if (! is_array($array) || ! is_array($keys)) return FALSE;
 158
 159         // Nondestructive test
 160         if (! $force)
 161                 foreach(array_keys($keys) as $from)
 162                         if (! isset($array[$from]))
 163                                 return FALSE;
 164
 165         foreach($keys as $from => $to) {
 166                 if ($from === $to) continue;
 167                 if (! $force || isset($array[$from])) {
 168                         $array[$to] = & $array[$from];
 169                         unset($array[$from]);
 170                 } else  {
 171                         $array[$to] = $default;
 172                 }
 173         }
 174
 175         return TRUE;
 176 }
 177
 178 // ---------------------
 179 // Spam-uri pickup
 180
 181 // Domain exposure callback (See spam_uri_pickup_preprocess())
 182 // http://victim.example.org/?foo+site:nasty.example.com+bar
 183 // => http://nasty.example.com/?refer=victim.example.org
 184 // NOTE: 'refer=' is not so good for (at this time).
 185 // Consider about using IP address of the victim, try to avoid that.
 186 function _preg_replace_callback_domain_exposure($matches = array())
 187 {
 188         $result = '';
 189
 190         // Preserve the victim URI as a complicity or ...
 191         if (isset($matches[5])) {
 192                 $result =
 193                         $matches[1] . '://' .   // scheme
 194                         $matches[2] . '/' .             // victim.example.org
 195                         $matches[3];                    // The rest of all (before victim)
 196         }
 197
 198         // Flipped URI
 199         if (isset($matches[4])) {
 200                 $result =
 201                         $matches[1] . '://' .   // scheme
 202                         $matches[4] .                   // nasty.example.com
 203                         '/?refer=' . strtolower($matches[2]) .  // victim.example.org
 204                         ' ' . $result;
 205         }
 206
 207         return $result;
 208 }
 209
 210 // Preprocess: rawurldecode() and adding space(s) and something
 211 // to detect/count some URIs _if possible_
 212 // NOTE: It's maybe danger to var_dump(result). [e.g. 'javascript:']
 213 // [OK] http://victim.example.org/go?http%3A%2F%2Fnasty.example.org
 214 // [OK] http://victim.example.org/http://nasty.example.org
 215 function spam_uri_pickup_preprocess($string = '')
 216 {
 217         if (! is_string($string)) return '';
 218
 219         $string = rawurldecode($string);
 220
 221         // Domain exposure (See _preg_replace_callback_domain_exposure())
 222         $string = preg_replace_callback(
 223                 array(
 224                         // Something Google: http://www.google.com/supported_domains
 225                         '#(http)://([a-z0-9.]+\.google\.[a-z]{2,3}(?:\.[a-z]{2})?)/' .
 226                         '([a-z0-9?=&.%_+-]+)' .         // ?query=foo+
 227                         '\bsite:([a-z0-9.%_-]+)' .      // site:nasty.example.com
 228                         //'()' .        // Preserve or remove?
 229                         '#i',
 230                 ),
 231                 '_preg_replace_callback_domain_exposure',
 232                 $string
 233         );
 234
 235         // URI exposure (uriuri => uri uri)
 236         $string = preg_replace(
 237                 array(
 238                         '#(?<! )(?:https?|ftp):/#i',
 239                 //      '#[a-z][a-z0-9.+-]{1,8}://#i',
 240                 //      '#[a-z][a-z0-9.+-]{1,8}://#i'
 241                 ),
 242                 ' $0',
 243                 $string
 244         );
 245
 246         return $string;
 247 }
 248
 249 // Main function of spam-uri pickup
 250 function spam_uri_pickup($string = '', $area = array())
 251 {
 252         if (! is_array($area) || empty($area)) {
 253                 $area = array(
 254                                 'anchor' => TRUE,
 255                                 'bbcode' => TRUE,
 256                         );
 257         }
 258
 259         $string = spam_uri_pickup_preprocess($string);
 260
 261         $array  = uri_pickup($string);
 262
 263         // Area elevation for '(especially external)link' intension
 264         if (! empty($array)) {
 265
 266                 $area_shadow = array();
 267                 foreach(array_keys($array) as $key){
 268                         $area_shadow[$key] = & $array[$key]['area'];
 269                         $area_shadow[$key]['anchor'] = 0;
 270                         $area_shadow[$key]['bbcode'] = 0;
 271                 }
 272
 273                 // Anchor tags by preg_match_all()
 274                 // [OK] <a href="http://nasty.example.com">visit http://nasty.example.com/</a>
 275                 // [OK] <a href=\'http://nasty.example.com/\' >discount foobar</a>
 276                 // [NG] <a href="http://ng.example.com">visit http://ng.example.com _not_ended_
 277                 // [NG] <a href=  >Good site!</a> <a href= "#" >test</a>
 278                 if (isset($area['anchor'])) {
 279                         $areas = array();
 280                         preg_match_all('#<a\b[^>]*href[^>]*>.*?</a\b[^>]*(>)#i',
 281                                  $string, $areas, PREG_SET_ORDER | PREG_OFFSET_CAPTURE);
 282                         //var_dump(recursive_map('htmlspecialchars', $areas));
 283                         foreach(array_keys($areas) as $_area) {
 284                                 $areas[$_area] =  array(
 285                                         $areas[$_area][0][1], // Area start (<a href>)
 286                                         $areas[$_area][1][1], // Area end   (</a>)
 287                                 );
 288                         }
 289                         area_measure($areas, $area_shadow, 1, 'anchor');
 290                 }
 291
 292                 // phpBB's "BBCode" by preg_match_all()
 293                 // [url]http://nasty.example.com/[/url]
 294                 // [link]http://nasty.example.com/[/link]
 295                 // [url=http://nasty.example.com]visit http://nasty.example.com/[/url]
 296                 // [link http://nasty.example.com/]buy something[/link]
 297                 // ?? [url=][/url]
 298                 if (isset($area['bbcode'])) {
 299                         $areas = array();
 300                         preg_match_all('#\[(url|link)\b[^\]]*\].*?\[/\1\b[^\]]*(\])#i',
 301                                  $string, $areas, PREG_SET_ORDER | PREG_OFFSET_CAPTURE);
 302                         //var_dump(recursive_map('htmlspecialchars', $areas));
 303                         foreach(array_keys($areas) as $_area) {
 304                                 $areas[$_area] = array(
 305                                         $areas[$_area][0][1], // Area start ([url])
 306                                         $areas[$_area][2][1], // Area end   ([/url])
 307                                 );
 308                         }
 309                         area_measure($areas, $area_shadow, 1, 'bbcode');
 310                 }
 311
 312                 // Various Wiki syntax
 313                 // [text_or_uri>text_or_uri]
 314                 // [text_or_uri:text_or_uri]
 315                 // [text_or_uri|text_or_uri]
 316                 // [text_or_uri->text_or_uri]
 317                 // [text_or_uri text_or_uri] // MediaWiki
 318                 // MediaWiki: [http://nasty.example.com/ visit http://nasty.example.com/]
 319
 320                 // Remove 'offset's for area_measure()
 321                 foreach(array_keys($array) as $key)
 322                         unset($array[$key]['area']['offset']);
 323         }
 324
 325         return $array;
 326 }
 327
 328 // If in doubt, it's a little doubtful
 329 function area_measure($areas, & $array, $belief = -1, $a_key = 'area', $o_key = 'offset')
 330 {
 331         if (! is_array($areas) || ! is_array($array)) return;
 332
 333         $areas_keys = array_keys($areas);
 334         foreach(array_keys($array) as $u_index) {
 335                 $offset = isset($array[$u_index][$o_key]) ?
 336                         intval($array[$u_index][$o_key]) : 0;
 337                 foreach($areas_keys as $a_index) {
 338                         if (isset($array[$u_index][$a_key])) {
 339                                 $offset_s = intval($areas[$a_index][0]);
 340                                 $offset_e = intval($areas[$a_index][1]);
 341                                 // [Area => inside <= Area]
 342                                 if ($offset_s < $offset && $offset < $offset_e) {
 343                                         $array[$u_index][$a_key] += $belief;
 344                                 }
 345                         }
 346                 }
 347         }
 348 }
 349
 350 // ---------------------
 351 // Normalization
 352
 353 // Scheme normalization: Renaming the schemes
 354 // snntp://example.org =>  nntps://example.org
 355 // NOTE: Keep the static lists simple. See also port_normalize().
 356 function scheme_normalize($scheme = '', $considerd_harmfull = TRUE)
 357 {
 358         // Abbreviations considerable they don't have link intension
 359         static $abbrevs = array(
 360                 'ttp'   => 'http',
 361                 'ttps'  => 'https',
 362         );
 363
 364         // Alias => normalized
 365         static $aliases = array(
 366                 'pop'   => 'pop3',
 367                 'news'  => 'nntp',
 368                 'imap4' => 'imap',
 369                 'snntp' => 'nntps',
 370                 'snews' => 'nntps',
 371                 'spop3' => 'pop3s',
 372                 'pops'  => 'pop3s',
 373         );
 374
 375         $scheme = strtolower(trim($scheme));
 376         if (isset($abbrevs[$scheme])) {
 377                 if ($considerd_harmfull) {
 378                         $scheme = $abbrevs[$scheme];
 379                 } else {
 380                         $scheme = '';
 381                 }
 382         }
 383         if (isset($aliases[$scheme])) $scheme = $aliases[$scheme];
 384
 385         return $scheme;
 386 }
 387
 388 // Port normalization: Suppress the (redundant) default port
 389 // HTTP://example.org:80/ => http://example.org/
 390 // HTTP://example.org:8080/ => http://example.org:8080/
 391 // HTTPS://example.org:443/ => https://example.org/
 392 function port_normalize($port, $scheme, $scheme_normalize = TRUE)
 393 {
 394         // Schemes that users _maybe_ want to add protocol-handlers
 395         // to their web browsers. (and attackers _maybe_ want to use ...)
 396         // Reference: http://www.iana.org/assignments/port-numbers
 397         static $array = array(
 398                 // scheme => default port
 399                 'ftp'     =>    21,
 400                 'ssh'     =>    22,
 401                 'telnet'  =>    23,
 402                 'smtp'    =>    25,
 403                 'tftp'    =>    69,
 404                 'gopher'  =>    70,
 405                 'finger'  =>    79,
 406                 'http'    =>    80,
 407                 'pop3'    =>   110,
 408                 'sftp'    =>   115,
 409                 'nntp'    =>   119,
 410                 'imap'    =>   143,
 411                 'irc'     =>   194,
 412                 'wais'    =>   210,
 413                 'https'   =>   443,
 414                 'nntps'   =>   563,
 415                 'rsync'   =>   873,
 416                 'ftps'    =>   990,
 417                 'telnets' =>   992,
 418                 'imaps'   =>   993,
 419                 'ircs'    =>   994,
 420                 'pop3s'   =>   995,
 421                 'mysql'   =>  3306,
 422         );
 423
 424         $port = trim($port);
 425         if ($port === '') return $port;
 426
 427         if ($scheme_normalize) $scheme = scheme_normalize($scheme);
 428         if (isset($array[$scheme]) && $port == $array[$scheme])
 429                 $port = ''; // Ignore the defaults
 430
 431         return $port;
 432 }
 433
 434 // Path normalization
 435 // http://example.org => http://example.org/
 436 // http://example.org#hoge => http://example.org/#hoge
 437 // http://example.org/path/a/b/./c////./d => http://example.org/path/a/b/c/d
 438 // http://example.org/path/../../a/../back => http://example.org/back
 439 function path_normalize($path = '', $divider = '/', $addroot = TRUE)
 440 {
 441         if (! is_string($path) || $path == '')
 442                 return $addroot ? $divider : '';
 443
 444         $path = trim($path);
 445         $last = ($path[strlen($path) - 1] == $divider) ? $divider : '';
 446         $array = explode($divider, $path);
 447
 448         // Remove paddings
 449         foreach(array_keys($array) as $key) {
 450                 if ($array[$key] == '' || $array[$key] == '.')
 451                          unset($array[$key]);
 452         }
 453         // Back-track
 454         $tmp = array();
 455         foreach($array as $value) {
 456                 if ($value == '..') {
 457                         array_pop($tmp);
 458                 } else {
 459                         array_push($tmp, $value);
 460                 }
 461         }
 462         $array = & $tmp;
 463
 464         $path = $addroot ? $divider : '';
 465         if (! empty($array)) $path .= implode($divider, $array) . $last;
 466
 467         return $path;
 468 }
 469
 470 // DirectoryIndex normalize (Destructive and rough)
 471 function file_normalize($string = 'index.html.en')
 472 {
 473         static $array = array(
 474                 'index'                 => TRUE,        // Some system can omit the suffix
 475                 'index.htm'             => TRUE,
 476                 'index.html'    => TRUE,
 477                 'index.shtml'   => TRUE,
 478                 'index.jsp'             => TRUE,
 479                 'index.php'             => TRUE,
 480                 'index.php3'    => TRUE,
 481                 'index.php4'    => TRUE,
 482                 //'index.pl'    => TRUE,
 483                 //'index.py'    => TRUE,
 484                 //'index.rb'    => TRUE,
 485                 'index.cgi'             => TRUE,
 486                 'default.htm'   => TRUE,
 487                 'default.html'  => TRUE,
 488                 'default.asp'   => TRUE,
 489                 'default.aspx'  => TRUE,
 490         );
 491
 492         // Content-negothiation filter:
 493         // Roughly removing ISO 639 -like
 494         // 2-letter suffixes (See RFC3066)
 495         $matches = array();
 496         if (preg_match('/(.*)\.[a-z][a-z](?:-[a-z][a-z])?$/i', $string, $matches)) {
 497                 $_string = $matches[1];
 498         } else {
 499                 $_string = & $string;
 500         }
 501
 502         if (isset($array[strtolower($_string)])) {
 503                 return '';
 504         } else {
 505                 return $string;
 506         }
 507 }
 508
 509 // Sort query-strings if possible (Destructive and rough)
 510 // [OK] &&&&f=d&b&d&c&a=0dd  =>  a=0dd&b&c&d&f=d
 511 // [OK] nothing==&eg=dummy&eg=padding&eg=foobar  =>  eg=foobar
 512 function query_normalize($string = '', $equal = FALSE, $equal_cutempty = TRUE)
 513 {
 514         $array = explode('&', $string);
 515
 516         // Remove '&' paddings
 517         foreach(array_keys($array) as $key) {
 518                 if ($array[$key] == '') {
 519                          unset($array[$key]);
 520                 }
 521         }
 522
 523         // Consider '='-sepalated input and paddings
 524         if ($equal) {
 525                 $equals = $not_equals = array();
 526                 foreach ($array as $part) {
 527                         if (strpos($part, '=') === FALSE) {
 528                                  $not_equals[] = $part;
 529                         } else {
 530                                 list($key, $value) = explode('=', $part, 2);
 531                                 $value = ltrim($value, '=');
 532                                 if (! $equal_cutempty || $value != '') {
 533                                         $equals[$key] = $value;
 534                                 }
 535                         }
 536                 }
 537
 538                 $array = & $not_equals;
 539                 foreach ($equals as $key => $value) {
 540                         $array[] = $key . '=' . $value;
 541                 }
 542                 unset($equals);
 543         }
 544
 545         natsort($array);
 546         return implode('&', $array);
 547 }
 548
 549 // ---------------------
 550 // Part One : Checker
 551
 552 function generate_glob_regex($string = '', $divider = '/')
 553 {
 554         static $from = array(
 555                          1 => '*',
 556                         11 => '?',
 557         //              22 => '[',      // Maybe cause regex compilation error (e.g. '[]')
 558         //              23 => ']',      //
 559                 );
 560         static $mid = array(
 561                          1 => '_AST_',
 562                         11 => '_QUE_',
 563         //              22 => '_RBR_',
 564         //              23 => '_LBR_',
 565                 );
 566         static $to = array(
 567                          1 => '.*',
 568                         11 => '.',
 569         //              22 => '[',
 570         //              23 => ']',
 571                 );
 572
 573         if (is_array($string)) {
 574                 // Recurse
 575                 return '(?:' .
 576                         implode('|',    // OR
 577                                 array_map('generate_glob_regex',
 578                                         $string,
 579                                         array_pad(array(), count($string), $divider)
 580                                 )
 581                         ) .
 582                 ')';
 583         } else {
 584                 $string = str_replace($from, $mid, $string); // Hide
 585                 $string = preg_quote($string, $divider);
 586                 $string = str_replace($mid, $to, $string);   // Unhide
 587                 return $string;
 588         }
 589 }
 590
 591 // TODO: Ignore list
 592 // TODO: preg_grep() ?
 593 // TODO: Multi list
 594 function is_badhost($hosts = '', $asap = TRUE)
 595 {
 596         static $regex;
 597
 598         if (! isset($regex)) {
 599                 $regex = array();
 600                 $regex['badhost'] = array();
 601
 602                 // Sample
 603                 if (TRUE) {
 604                         $blocklist['badhost'] = array(
 605                                 //'*',                  // Deny all uri
 606                                 //'10.20.*.*',  // 10.20.example.com also matches
 607                                 //'*.blogspot.com',     // Blog services subdomains
 608                                 //array('blogspot.com', '*.blogspot.com')
 609                         );
 610                         foreach ($blocklist['badhost'] as $part) {
 611                                 $regex['badhost'][] = '/^' . generate_glob_regex($part) . '$/i';
 612                         }
 613                 }
 614
 615                 // Load
 616                 if (file_exists(SPAM_INI_FILE)) {
 617                         $blocklist = array();
 618                         require(SPAM_INI_FILE);
 619                         foreach ($blocklist['badhost'] as $part) {
 620                                 $regex['badhost'][] = '/^' . generate_glob_regex($part) . '$/i';
 621                         }
 622                 }
 623         }
 624         //var_dump($regex);
 625
 626         $result = 0;
 627         if (! is_array($hosts)) $hosts = array($hosts);
 628
 629         foreach($hosts as $host) {
 630                 if (! is_string($host)) $host = '';
 631
 632                 // badhost
 633                 foreach ($regex['badhost'] as $_regex) {
 634                         if (preg_match($_regex, $host)) {
 635                                 ++$result;
 636                                 if ($asap) {
 637                                         return $result;
 638                                 } else {
 639                                         break;
 640                                 }
 641                         }
 642                 }
 643         }
 644
 645         return $result;
 646 }
 647
 648 // Default (enabled) methods and thresholds
 649 function check_uri_spam_method($times = 1, $t_area = 0, $rule = TRUE)
 650 {
 651         $times  = intval($times);
 652         $t_area = intval($t_area);
 653
 654         // Thresholds
 655         $method = array(
 656                 'quantity' => 8 * $times,       // Allow N URIs
 657                 'non_uniq' => 3 * $times,       // Allow N duped (and normalized) URIs
 658         );
 659
 660         // Areas
 661         $area = array(
 662                 //'total'  => $t_area,  // Allow N areas total, enabled below
 663                 'anchor'   => $t_area,  // Inside <a href> HTML tag
 664                 'bbcode'   => $t_area,  // Inside [url] or [link] BBCode
 665         );
 666
 667         // Rules
 668         $rules = array(
 669                 'badhost'  => TRUE,     // Check badhost
 670         );
 671
 672         // Remove unused
 673         foreach (array_keys($method) as $key) {
 674                 if ($method[$key] < 0) unset($method[$key]);
 675         }
 676         foreach (array_keys($area) as $key) {
 677                 if ($area[$key] < 0) unset($area[$key]);
 678         }
 679         $area  = empty($area) ? array() : array('area' => $area);
 680         $rules = $rule ? $rules : array();
 681
 682         return $method + $area + $rules;
 683 }
 684
 685 // TODO: Simplify $progress data structure
 686 // TODO: Simplify. !empty(['_action']) just means $is_spam
 687 // Simple/fast spam check
 688 function check_uri_spam($target = '', $method = array(), $asap = TRUE)
 689 {
 690         $is_spam  = FALSE;
 691         $progress = array(
 692                 'quantity'   => 0,
 693                 'area' => array(
 694                         'total'  => 0,
 695                         'anchor' => 0,
 696                         'bbcode' => 0,
 697                         ),
 698                 'non_uniq'   => 0,
 699                 'uniqhost'   => 0,
 700                 'badhost'    => 0,
 701                 '_action'    => array(),
 702                 );
 703
 704         if (! is_array($method) || empty($method)) {
 705                 $method = check_uri_spam_method();
 706         }
 707
 708         if (is_array($target)) {
 709                 // Recurse
 710                 foreach($target as $str) {
 711                         list($_is_spam, $_progress) = check_uri_spam($str, $method, $asap);
 712                         $is_spam = $is_spam || $_is_spam;
 713                         $progress['quantity']       += $_progress['quantity'];
 714                         $progress['area']['total']  += $_progress['area']['total'];
 715                         $progress['area']['anchor'] += $_progress['area']['anchor'];
 716                         $progress['area']['bbcode'] += $_progress['area']['bbcode'];
 717                         $progress['non_uniq']       += $_progress['non_uniq'];
 718                         $progress['uniqhost']       += $_progress['uniqhost'];
 719                         $progress['badhost']        += $_progress['badhost'];
 720                         foreach($_progress['_action'] as $key => $value) {
 721                                 if (is_array($value)) {
 722                                         foreach(array_keys($value) as $_key) {
 723                                                 $progress['_action'][$key][$_key] = TRUE;
 724                                         }
 725                                 } else {
 726                                         $progress['_action'][$key] = TRUE;
 727                                 }
 728                         }
 729                         if ($is_spam && $asap) break;
 730                 }
 731         } else {
 732                 $pickups = spam_uri_pickup($target);
 733                 if (! empty($pickups)) {
 734                         $progress['quantity'] += count($pickups);
 735
 736                         // URI quantity
 737                         if ((! $is_spam || ! $asap) && isset($method['quantity']) &&
 738                                 $progress['quantity'] > $method['quantity']) {
 739                                 $is_spam = TRUE;
 740                                 $progress['_action']['quantity'] = TRUE;
 741                         }
 742                         //var_dump($method['quantity'], $is_spam);
 743
 744                         // Using invalid area
 745                         if ((! $is_spam || ! $asap) && isset($method['area'])) {
 746                                 foreach($pickups as $pickup) {
 747                                         foreach ($pickup['area'] as $key => $value) {
 748                                                 if ($key == 'offset') continue;
 749                                                 $progress['area']['total'] += $value;
 750                                                 $progress['area'][$key]    += $value;
 751                                                 if (isset($method['area']['total']) &&
 752                                                                 $progress['area']['total'] > $method['area']['total']) {
 753                                                         $is_spam = TRUE;
 754                                                         $progress['_action']['area']['total'] = TRUE;
 755                                                         if ($is_spam && $asap) break;
 756                                                 }
 757                                                 if(isset($method['area'][$key]) &&
 758                                                                 $progress['area'][$key] > $method['area'][$key]) {
 759                                                         $is_spam = TRUE;
 760                                                         $progress['_action']['area'][$key] = TRUE;
 761                                                         if ($is_spam && $asap) break;
 762                                                 }
 763                                         }
 764                                         if ($is_spam && $asap) break;
 765                                 }
 766                         }
 767                         //var_dump($method['area'], $is_spam);
 768
 769
 770                         // URI uniqueness (and removing non-uniques)
 771                         if ((! $is_spam || ! $asap) && isset($method['non_uniq'])) {
 772
 773                                 // Destructive normalize of URIs
 774                                 uri_array_normalize($pickups);
 775
 776                                 $uris = array();
 777                                 foreach (array_keys($pickups) as $key) {
 778                                         $uris[$key] = uri_array_implode($pickups[$key]);
 779                                 }
 780                                 $count = count($uris);
 781                                 $uris  = array_unique($uris);
 782                                 $progress['non_uniq'] += $count - count($uris);
 783                                 if ($progress['non_uniq'] > $method['non_uniq']) {
 784                                         $is_spam = TRUE;
 785                                         $progress['_action']['non_uniq'] = TRUE;
 786                                 }
 787                                 if (! $asap || ! $is_spam) {
 788                                         foreach (array_diff(array_keys($pickups),
 789                                                 array_keys($uris)) as $remove) {
 790                                                 unset($pickups[$remove]);
 791                                         }
 792                                 }
 793                                 unset($uris);
 794                         }
 795                         //var_dump($method['non_uniq'], $is_spam);
 796
 797                         // Unique host
 798                         $hosts = array();
 799                         foreach ($pickups as $pickup) {
 800                                 $hosts[] = & $pickup['host'];
 801                         }
 802                         $hosts = array_unique($hosts);
 803                         $progress['uniqhost'] += count($hosts);
 804                         //var_dump($method['uniqhost'], $is_spam);
 805
 806                         // Bad host
 807                         if ((! $is_spam || ! $asap) && isset($method['badhost'])) {
 808                                 $count = is_badhost($hosts, $asap);
 809                                 $progress['badhost'] += $count;
 810                                 if ($count !== 0) {
 811                                         $progress['_action']['badhost'] = TRUE;
 812                                         $is_spam = TRUE;
 813                                 }
 814                         }
 815                         //var_dump($method['badhost'], $is_spam);
 816                 }
 817         }
 818
 819         return array($is_spam, $progress);
 820 }
 821
 822 // ---------------------
 823 // Reporting
 824
 825 // TODO: Show all
 826 // Summarize $progress (blocked only)
 827 function summarize_check_uri_spam_progress($progress = array(), $shownum = TRUE)
 828 {
 829         //$list = check_uri_spam_method();
 830
 831         $tmp = array();
 832         foreach (array_keys($progress['_action']) as $_action) {
 833                 if (is_array($progress['_action'][$_action])) {
 834                         foreach (array_keys($progress['_action'][$_action]) as $_area) {
 835                                 $tmp[] = $_action . '=>' . $_area .
 836                                         ($shownum ? '(' . $progress[$_action][$_area] . ')' : '');
 837                         }
 838                 } else {
 839                         $tmp[] = $_action .
 840                                         ($shownum ? '(' . $progress[$_action] . ')' : '');
 841                 }
 842         }
 843
 844         return implode(', ', $tmp);
 845 }
 846
 847 // ---------------------
 848 // Exit
 849
 850 // Common bahavior for blocking
 851 // NOTE: Call this function from various blocking feature, to disgueise the reason 'why blocked'
 852 function spam_exit()
 853 {
 854         die("\n");
 855 }
 856
 857
 858 // ---------------------
 859 // Simple filtering
 860
 861 // TODO: Record them
 862 // Simple/fast spam filter ($target: 'a string' or an array())
 863 function pkwk_spamfilter($action, $page, $target = array('title' => ''), $method = array(), $asap = FALSE)
 864 {
 865         global $notify;
 866
 867         list($is_spam, $progress) = check_uri_spam($target, $method, $asap);
 868
 869         if ($is_spam) {
 870                 // Mail to administrator(s)
 871                 if ($notify) pkwk_spamnotify($action, $page, $target, $progress);
 872                 // End
 873                 spam_exit();
 874         }
 875 }
 876
 877 // ---------------------
 878 // PukiWiki original
 879
 880 // Mail to administrator(s)
 881 function pkwk_spamnotify($action, $page, $target = array('title' => ''), $progress = array())
 882 {
 883         global $notify_subject;
 884
 885         $footer['ACTION'] = 'Blocked by: ' . summarize_check_uri_spam_progress($progress);
 886         $footer['COMMENT'] = $action;
 887         $footer['PAGE']   = '[blocked] ' . $page;
 888         $footer['URI']    = get_script_uri() . '?' . rawurlencode($page);
 889         $footer['USER_AGENT']  = TRUE;
 890         $footer['REMOTE_ADDR'] = TRUE;
 891         pkwk_mail_notify($notify_subject,  var_export($target, TRUE), $footer);
 892 }
 893
 894 ?>