| 1 |
<?php |
| 2 |
|
| 3 |
if (!defined('WEBTOTEM_INIT') || WEBTOTEM_INIT !== true) { |
| 4 |
if (!headers_sent()) { |
| 5 |
header('HTTP/1.1 403 Forbidden'); |
| 6 |
} |
| 7 |
die("Protected By WebTotem!"); |
| 8 |
} |
| 9 |
|
| 10 |
/** |
| 11 |
* WebTotem page scan class for Wordpress. |
| 12 |
*/ |
| 13 |
class WebTotemCrawler |
| 14 |
{ |
| 15 |
/** |
| 16 |
* Running a single iteration |
| 17 |
* |
| 18 |
* @param array $scan_temp |
| 19 |
* The data of the current scan. |
| 20 |
*/ |
| 21 |
public static function init($scan_temp) { |
| 22 |
|
| 23 |
$crawler_temp = json_decode(WebTotemOption::getOption('crawler_temp'), true) ?: []; |
| 24 |
|
| 25 |
$i = 1; |
| 26 |
if (!$crawler_temp) { |
| 27 |
$pre_scan = self::pre_scan(); |
| 28 |
|
| 29 |
$crawler_temp['internal']['new'] = $pre_scan['internal']; |
| 30 |
$crawler_temp['external'] = WebTotem::arrayUniqueKey($pre_scan['external'], 'link'); |
| 31 |
$crawler_temp['scripts'] = WebTotem::arrayUniqueKey($pre_scan['scripts'], 'link'); |
| 32 |
$crawler_temp['iframes'] = WebTotem::arrayUniqueKey( $pre_scan['iframes'], 'link'); |
| 33 |
$crawler_temp['exclude'] = $pre_scan['exclude']; |
| 34 |
|
| 35 |
$crawler_temp['internal']['new'] = WebTotem::arrayUniqueKey( array_merge($crawler_temp['internal']['new'], $scan_temp['links']), 'link'); |
| 36 |
|
| 37 |
$i++; |
| 38 |
} |
| 39 |
|
| 40 |
$visited = []; |
| 41 |
|
| 42 |
foreach ($crawler_temp['internal']['new'] as $key => $item) { |
| 43 |
if($result = self::explore_page($item['link'], $crawler_temp['exclude'])) { |
| 44 |
|
| 45 |
$crawler_temp['internal']['visited'][] = $item; |
| 46 |
|
| 47 |
$crawler_temp['internal']['new'] = WebTotem::arrayUniqueKey(array_merge($crawler_temp['internal']['new'] ?? [], $result['internal'] ?? []), 'link'); |
| 48 |
$crawler_temp['external'] = WebTotem::arrayUniqueKey(array_merge($crawler_temp['external'] ?? [], $result['external'] ?? []), 'link'); |
| 49 |
$crawler_temp['scripts'] = WebTotem::arrayUniqueKey(array_merge($crawler_temp['scripts'] ?? [], $result['scripts'] ?? []), 'link'); |
| 50 |
$crawler_temp['iframes'] = WebTotem::arrayUniqueKey(array_merge($crawler_temp['iframes'] ?? [], $result['iframes'] ?? []), 'link'); |
| 51 |
$crawler_temp['exclude'] = array_merge($crawler_temp['exclude'] ?? [], $result['exclude'] ?? []); |
| 52 |
} |
| 53 |
|
| 54 |
$visited[] = $key; |
| 55 |
|
| 56 |
if ($i >= 5) break; |
| 57 |
$i++; |
| 58 |
} |
| 59 |
|
| 60 |
foreach ($visited as $key){ |
| 61 |
unset($crawler_temp['internal']['new'][$key]); |
| 62 |
} |
| 63 |
WebTotemOption::setOptions(['crawler_temp' => $crawler_temp]); |
| 64 |
|
| 65 |
if (empty($crawler_temp['internal']['new'])) { |
| 66 |
|
| 67 |
if($scan_temp['ready_to_save']){ |
| 68 |
|
| 69 |
if(isset($crawler_temp['internal']['visited']) and $crawler_temp['external']){ |
| 70 |
$links = array_merge($crawler_temp['internal']['visited'], $crawler_temp['external']); |
| 71 |
} elseif (isset($crawler_temp['internal']['visited'])){ |
| 72 |
$links = $crawler_temp['internal']['visited']; |
| 73 |
} else { |
| 74 |
$links = $crawler_temp['external']; |
| 75 |
} |
| 76 |
|
| 77 |
$data = [ |
| 78 |
'links' => $links ?? [], |
| 79 |
'scripts' => $crawler_temp['scripts'] ?? [], |
| 80 |
'iframes' => $crawler_temp['iframes'] ?? [], |
| 81 |
]; |
| 82 |
|
| 83 |
self::saveData($data); |
| 84 |
|
| 85 |
WebTotemOption::setOptions(['crawler_temp' => '']); |
| 86 |
WebTotemOption::setOptions(['scan_temp' => '']); |
| 87 |
WebTotemOption::setOptions(['scan_init' => 0]); |
| 88 |
|
| 89 |
// Resetting the task in the cron. |
| 90 |
wp_clear_scheduled_hook('webtotem_daily_cron'); |
| 91 |
wp_schedule_event(time() + 86395, 'daily', 'webtotem_daily_cron'); |
| 92 |
} else { |
| 93 |
WebTotemOption::setOptions([ |
| 94 |
'scan_temp' => [ |
| 95 |
'current_scan' => 'crawler', |
| 96 |
'links' => [], |
| 97 |
'ready_to_save' => true, |
| 98 |
] |
| 99 |
]); |
| 100 |
} |
| 101 |
|
| 102 |
} |
| 103 |
|
| 104 |
} |
| 105 |
|
| 106 |
/** |
| 107 |
* Preliminary scan of the site. |
| 108 |
* |
| 109 |
* @return array |
| 110 |
*/ |
| 111 |
private static function pre_scan() { |
| 112 |
$site_url = get_site_url(); |
| 113 |
$internal = []; |
| 114 |
$exclude = [$site_url]; |
| 115 |
|
| 116 |
// Scanning the file robots.txt |
| 117 |
if(file_exists(ABSPATH . 'robots.txt')){ |
| 118 |
$robotsTxt = file_get_contents(ABSPATH . 'robots.txt'); |
| 119 |
$lines = explode("\n", $robotsTxt); |
| 120 |
$robots_urls = []; |
| 121 |
|
| 122 |
foreach ($lines as $line) { |
| 123 |
if (strpos($line, 'Disallow:') === 0 || strpos($line, 'Allow:') === 0) { |
| 124 |
$url = trim(substr($line, strpos($line, ':') + 1)); |
| 125 |
$exclude[] = $url; |
| 126 |
$robots_urls[] = (string)$url->loc; |
| 127 |
} |
| 128 |
} |
| 129 |
|
| 130 |
foreach ($robots_urls as $link) { |
| 131 |
if (substr($link, 0, 1) == "#") { |
| 132 |
continue; |
| 133 |
} |
| 134 |
$link = untrailingslashit($link); |
| 135 |
$internal[] = ['link' => $link, 'page' => $site_url . '/robots.txt', 'is_internal' => self::isInternal($link)]; |
| 136 |
} |
| 137 |
} |
| 138 |
|
| 139 |
|
| 140 |
// Adding links from popular sitemaps plugins |
| 141 |
$sitemaps = [ |
| 142 |
$site_url . '/sitemaps.xml', |
| 143 |
$site_url . '/index.php?xml_sitemap=params=.', |
| 144 |
$site_url . '/?sitemap=1', |
| 145 |
$site_url . '/sitemap_index.xml', |
| 146 |
]; |
| 147 |
|
| 148 |
foreach ($sitemaps as $link) { |
| 149 |
$internal[] = ['link' => $link, 'page' => __('by sitemap plugins', 'wtotem'), 'is_internal' => true]; |
| 150 |
} |
| 151 |
|
| 152 |
// Scanning the file sitemap.xml |
| 153 |
$xml = simplexml_load_file(ABSPATH . 'sitemap.xml'); |
| 154 |
|
| 155 |
if($xml){ |
| 156 |
$sitemap_urls = []; |
| 157 |
foreach ($xml->url as $url) { |
| 158 |
$exclude[] = (string)$url->loc; |
| 159 |
$sitemap_urls[] = (string)$url->loc; |
| 160 |
} |
| 161 |
|
| 162 |
foreach ($sitemap_urls as $link) { |
| 163 |
$link = untrailingslashit($link); |
| 164 |
if (substr($link, 0, 1) == "#") { |
| 165 |
continue; |
| 166 |
} |
| 167 |
$internal[] = ['link' => $link, 'page' => $site_url . '/sitemap.xml', 'is_internal' => self::isInternal($link)]; |
| 168 |
} |
| 169 |
} |
| 170 |
|
| 171 |
// Scanning the main page |
| 172 |
$result = self::explore_page($site_url, $exclude); |
| 173 |
|
| 174 |
$internal = array_merge($internal, $result['internal']); |
| 175 |
$external = array_unique($result['external']); |
| 176 |
$exclude = array_merge($exclude, $result['exclude'] ?? []); |
| 177 |
|
| 178 |
return [ |
| 179 |
'internal' => $internal ?: [], |
| 180 |
'external' => $external ?: [], |
| 181 |
'scripts' => $result['scripts'] ?: [], |
| 182 |
'iframes' => $result['iframes'] ?: [], |
| 183 |
'exclude' => $exclude ?: [], |
| 184 |
]; |
| 185 |
|
| 186 |
} |
| 187 |
|
| 188 |
/** |
| 189 |
* Get and explore the content of the page. |
| 190 |
* |
| 191 |
* @param string $url |
| 192 |
* Link to the page. |
| 193 |
* @param string $exclude |
| 194 |
* Links that have already been checked. |
| 195 |
* |
| 196 |
* @return array|bool |
| 197 |
*/ |
| 198 |
private static function explore_page($url, $exclude = []) { |
| 199 |
|
| 200 |
if(!$url){ |
| 201 |
return false; |
| 202 |
} |
| 203 |
|
| 204 |
$headers = get_headers($url); |
| 205 |
|
| 206 |
if ($headers === false || strpos($headers[0], '200 OK') === false) { |
| 207 |
return false; |
| 208 |
} |
| 209 |
|
| 210 |
// Initializing the cURL session |
| 211 |
$curl = curl_init(); |
| 212 |
|
| 213 |
// Setting the parameters of the cURL session |
| 214 |
curl_setopt($curl, CURLOPT_URL, $url); // Setting the URL |
| 215 |
curl_setopt($curl, CURLOPT_RETURNTRANSFER, true); // We return the result as a string |
| 216 |
curl_setopt($curl, CURLOPT_FOLLOWLOCATION, true); // Follow the redirects |
| 217 |
curl_setopt($curl, CURLOPT_SSL_VERIFYPEER, false); // Disabling SSL certificate verification |
| 218 |
|
| 219 |
// Execute the request and get the content of the page |
| 220 |
$content = curl_exec($curl); |
| 221 |
|
| 222 |
// Checking for errors when executing the request |
| 223 |
if (curl_errno($curl)) { |
| 224 |
WebTotemOption::setNotification('error', __('Request execution error: ', 'wtotem')) . curl_error($curl); |
| 225 |
} |
| 226 |
|
| 227 |
// Closing the cURL session |
| 228 |
curl_close($curl); |
| 229 |
|
| 230 |
// Checking the content for matches with the template, using regular expressions |
| 231 |
return self::getMatches($content, $url, $exclude); |
| 232 |
} |
| 233 |
|
| 234 |
/** |
| 235 |
* We are looking for links, scripts and frames on the page. |
| 236 |
* |
| 237 |
* @param string $content |
| 238 |
* The content of the page being checked. |
| 239 |
* @param string $url |
| 240 |
* Link to the page. |
| 241 |
* @param array $exclude |
| 242 |
* Links that have already been checked. |
| 243 |
* |
| 244 |
* @return array |
| 245 |
*/ |
| 246 |
private static function getMatches($content, $url, $exclude) { |
| 247 |
|
| 248 |
$matches = [ |
| 249 |
'internal' => [], |
| 250 |
'external' => [], |
| 251 |
'exclude' => [], |
| 252 |
'scripts' => [], |
| 253 |
'iframes' => [], |
| 254 |
]; |
| 255 |
|
| 256 |
if($content){ |
| 257 |
// Get all the matches. |
| 258 |
$pattern = '/(<a.*?href=["\']([^"\']+)["\'].*?>|<script.*?src=["\']([^"\']+)["\'].*?>|<iframe.*?src=["\']([^"\']+)["\'].*?>|onclick=["\']?[^"\']*location.*?["\']?([^"\']+))/i'; |
| 259 |
preg_match_all($pattern, $content, $all_matches); |
| 260 |
|
| 261 |
$array = [ |
| 262 |
'links' => [], |
| 263 |
'scripts' => [], |
| 264 |
'iframes' => [], |
| 265 |
]; |
| 266 |
|
| 267 |
// Divide by categories. |
| 268 |
foreach ($all_matches[0] as $match) { |
| 269 |
preg_match_all('/<a.*?href=(["](.*?)["]|[\'](.*?)[\']).*?>/i', $match, $links_matches); |
| 270 |
if (isset($links_matches[2])) $array['links'] = array_merge($array['links'], $links_matches[2]); |
| 271 |
preg_match_all('/onclick="[^"]*location[^"][^\'"]+\'([^\']+)\'/i', $match, $links_2_matches); |
| 272 |
if (isset($links_2_matches[2])) $array['links'] = array_merge($array['links'], $links_2_matches[2]); |
| 273 |
preg_match_all('/<script.*?src=(["](.*?)["]|[\'](.*?)[\']).*?>/i', $match, $js_matches); |
| 274 |
if (isset($js_matches[2])) $array['scripts'] = array_merge($array['scripts'], $js_matches[2]); |
| 275 |
preg_match_all('/<iframe.*?src=(["](.*?)["]|[\'](.*?)[\']).*?>/i', $match, $iframe_matches); |
| 276 |
if (isset($iframe_matches[2])) $array['iframes'] = array_merge($array['iframes'], $iframe_matches[2]); |
| 277 |
} |
| 278 |
|
| 279 |
foreach ($array['links'] as $link) { |
| 280 |
if($link){ |
| 281 |
if (self::isInternal($link)) { |
| 282 |
if (substr($link, 0, 1) != "#" and !in_array($link, $exclude)) { |
| 283 |
$matches['internal'][] = ['link' => $link, 'page' => $url, 'is_internal' => true]; |
| 284 |
$matches['exclude'][] = $link; |
| 285 |
$exclude[] = $link; |
| 286 |
} |
| 287 |
} else { |
| 288 |
if(!in_array($link, $exclude)){ |
| 289 |
$matches['external'][] = ['link' => $link, 'page' => $url, 'is_internal' => false]; |
| 290 |
$matches['exclude'][] = $link; |
| 291 |
$exclude[] = $link; |
| 292 |
} |
| 293 |
} |
| 294 |
} |
| 295 |
|
| 296 |
} |
| 297 |
|
| 298 |
foreach (array_unique($array['scripts']) as $script) { |
| 299 |
if($script){ |
| 300 |
$matches['scripts'][] = ['link' => $script, 'page' => $url, 'is_internal' => self::isInternal($script)]; |
| 301 |
} |
| 302 |
} |
| 303 |
foreach (array_unique($array['iframes']) as $iframe) { |
| 304 |
if($iframe){ |
| 305 |
$matches['iframes'][] = ['link' => $iframe, 'page' => $url, 'is_internal' => self::isInternal($iframe)]; |
| 306 |
} |
| 307 |
} |
| 308 |
|
| 309 |
} |
| 310 |
|
| 311 |
return $matches; |
| 312 |
} |
| 313 |
|
| 314 |
/** |
| 315 |
* We check whether the link is internal or external. |
| 316 |
* |
| 317 |
* @param string $string |
| 318 |
* The link being checked. |
| 319 |
* |
| 320 |
* @return bool |
| 321 |
*/ |
| 322 |
private static function isInternal($string): bool { |
| 323 |
$current_domain_parts = parse_url(get_home_url()); |
| 324 |
$current_domain = $current_domain_parts['host']; |
| 325 |
|
| 326 |
if (substr($string, 0, 5) == "https" |
| 327 |
|| substr($string, 0, 4) == "http" |
| 328 |
|| substr($string, 0, 2) == "//") { |
| 329 |
|
| 330 |
if (strpos($string, $current_domain) === false) { |
| 331 |
return false; |
| 332 |
} |
| 333 |
} |
| 334 |
|
| 335 |
return true; |
| 336 |
} |
| 337 |
|
| 338 |
/** |
| 339 |
* Save data. |
| 340 |
* |
| 341 |
* @param array $data |
| 342 |
* Array matches data. |
| 343 |
*/ |
| 344 |
private static function saveData($data) { |
| 345 |
|
| 346 |
WebTotemDB::deleteData([], 'scan_logs'); |
| 347 |
$values = ''; |
| 348 |
foreach ($data as $data_type => $links) { |
| 349 |
foreach ($links as $datum) { |
| 350 |
$values .= sprintf("('%s','%s','%s','%s','%s'),", |
| 351 |
date("Y-m-d H:i:s"), |
| 352 |
$data_type, |
| 353 |
$datum['page'], |
| 354 |
addslashes($datum['link']), |
| 355 |
$datum['is_internal'] |
| 356 |
); |
| 357 |
} |
| 358 |
} |
| 359 |
|
| 360 |
$values = substr_replace($values, ";", -1); |
| 361 |
|
| 362 |
$columns = '(created_at, data_type, source, content, is_internal)'; |
| 363 |
|
| 364 |
WebTotemDB::setRows('scan_logs', $columns, $values); |
| 365 |
} |
| 366 |
|
| 367 |
} |
| 368 |
|
