friendica-addons/blockbot/blockbot.php

85 lines
3.1 KiB
PHP
Raw Normal View History

2019-04-20 12:15:45 +00:00
<?php
/**
* Name: blockbot
2019-04-20 12:15:45 +00:00
* Description: Blocking bots based on detecting bots/crawlers/spiders via the user agent and http_from header.
* Version: 0.1
* Author: Philipp Holzer <admin@philipp.info>
*
*/
use Friendica\App;
2019-05-29 18:51:07 +00:00
use Friendica\Core\Config;
2019-04-20 12:15:45 +00:00
use Friendica\Core\Hook;
use Friendica\Core\System;
use Jaybizzle\CrawlerDetect\CrawlerDetect;
2019-04-27 11:51:44 +00:00
use Friendica\Core\Logger;
2019-04-20 12:15:45 +00:00
2019-04-21 10:35:33 +00:00
require_once __DIR__ . DIRECTORY_SEPARATOR . 'vendor' . DIRECTORY_SEPARATOR . 'autoload.php';
function blockbot_install() {
2019-04-22 08:49:40 +00:00
Hook::register('init_1', __FILE__, 'blockbot_init_1');
2019-04-20 12:15:45 +00:00
}
function blockbot_uninstall() {
2019-04-22 08:49:40 +00:00
Hook::unregister('init_1', __FILE__, 'blockbot_init_1');
2019-04-20 12:15:45 +00:00
}
function blockbot_init_1(App $a) {
2019-05-03 10:25:13 +00:00
if (empty($_SERVER['HTTP_USER_AGENT'])) {
return;
}
$logdata = ['agent' => $_SERVER['HTTP_USER_AGENT'], 'uri' => $_SERVER['REQUEST_URI']];
2019-05-30 04:45:20 +00:00
// List of known crawlers.
2019-05-29 18:51:07 +00:00
$agents = ['SEMrushBot', 's~feedly-nikon3', 'Qwantify/Bleriot/', 'ltx71', 'Sogou web spider/',
'Diffbot/', 'Twitterbot/', 'YisouSpider/', 'evc-batch/', 'LivelapBot/', 'TrendsmapResolver/',
'PaperLiBot/', 'Nuzzel', 'um-LN/', 'Google Favicon', 'Datanyze', 'BLEXBot/', '360Spider',
'adscanner/', 'HeadlessChrome', 'wpif', 'startmebot/', 'Googlebot/', 'Applebot/',
2019-05-29 19:13:15 +00:00
'facebookexternalhit/', 'GoogleImageProxy', 'bingbot/', 'heritrix/', 'ldspider',
2019-05-31 15:37:33 +00:00
'AwarioRssBot/', 'Zabbix', 'TweetmemeBot/', 'dcrawl/', 'PhantomJS/', 'Googlebot-Image/',
2019-06-01 04:51:01 +00:00
'CrowdTanglebot/', 'Mediapartners-Google', 'Baiduspider/', 'datagnionbot',
2019-06-06 20:31:16 +00:00
'MegaIndex.ru/', 'SMUrlExpander', 'Hatena-Favicon/', 'Wappalyzer', 'FlipboardProxy/',
'NetcraftSurveyAgent/', 'Dataprovider.com', 'SMTBot/', 'Nimbostratus-Bot/',
2019-06-10 14:33:42 +00:00
'DuckDuckGo-Favicons-Bot/', 'IndieWebCards/', 'proximic', 'netEstate NE Crawler',
2019-06-11 10:12:39 +00:00
'AhrefsBot/', 'YandexBot/', 'Exabot/', 'Mediumbot-MetaTagFetcher/',
2019-06-12 09:11:54 +00:00
'WhatsApp/', 'TelegramBot', 'SurdotlyBot/', 'BingPreview/', 'SabsimBot/'];
2019-05-29 18:51:07 +00:00
foreach ($agents as $agent) {
if (stristr($_SERVER['HTTP_USER_AGENT'], $agent)) {
System::httpExit(403, 'Bots are not allowed');
}
}
2019-05-30 04:45:20 +00:00
// This switch here is only meant for developers who want to add more bots to the list above, it is not safe for production.
2019-05-29 18:51:07 +00:00
if (!Config::get('blockbot', 'training')) {
return;
}
$crawlerDetect = new CrawlerDetect();
if (!$crawlerDetect->isCrawler()) {
logger::debug('Good user agent detected', $logdata);
return;
}
// List of false positives' strings of known "good" agents.
$agents = ['fediverse.network crawler', 'Active_Pods_CheckBot_3.0', 'Social-Relay/',
2019-04-28 06:44:56 +00:00
'curl', 'zgrab', 'Go-http-client', 'curb', 'github.com', 'reqwest', 'Feedly/',
'Python-urllib/', 'Liferea/', 'aiohttp/', 'WordPress.com Reader', 'hackney/',
'Faraday v', 'okhttp', 'UniversalFeedParser', 'PixelFedBot', 'python-requests',
2019-05-30 10:32:01 +00:00
'WordPress/', 'http.rb/', 'Apache-HttpClient/', 'WordPress.com;', 'Pleroma',
2019-06-12 09:07:30 +00:00
'Dispatch/', 'Ruby', 'Uptimebot/', 'Java/', 'libwww-perl/'];
foreach ($agents as $agent) {
if (stristr($_SERVER['HTTP_USER_AGENT'], $agent)) {
logger::notice('False positive', $logdata);
return;
}
}
logger::info('Blocked bot', $logdata);
2019-05-29 18:51:07 +00:00
System::httpExit(403, 'Bots are not allowed');
2019-04-20 12:15:45 +00:00
}