friendica-addons/blockbot/blockbot.php

133 lines
5.8 KiB
PHP
Raw Normal View History

2019-04-20 14:15:45 +02:00
<?php
/**
* Name: blockbot
2019-04-20 14:15:45 +02:00
* Description: Blocking bots based on detecting bots/crawlers/spiders via the user agent and http_from header.
* Version: 0.2
2019-04-20 14:15:45 +02:00
* Author: Philipp Holzer <admin@philipp.info>
* Author: Michael Vogel <https://pirati.ca/profile/heluecht>
2019-04-20 14:15:45 +02:00
*
*/
use Friendica\App;
use Friendica\Core\Hook;
use Friendica\Core\System;
2020-01-18 22:07:06 +01:00
use Friendica\DI;
2019-04-20 14:15:45 +02:00
use Jaybizzle\CrawlerDetect\CrawlerDetect;
2019-04-27 13:51:44 +02:00
use Friendica\Core\Logger;
use Friendica\Core\Renderer;
2019-04-20 14:15:45 +02:00
2019-04-21 12:35:33 +02:00
require_once __DIR__ . DIRECTORY_SEPARATOR . 'vendor' . DIRECTORY_SEPARATOR . 'autoload.php';
function blockbot_install() {
2019-04-22 10:49:40 +02:00
Hook::register('init_1', __FILE__, 'blockbot_init_1');
2019-04-20 14:15:45 +02:00
}
function blockbot_addon_admin(&$a, &$o) {
$t = Renderer::getMarkupTemplate("admin.tpl", "addon/blockbot/");
$o = Renderer::replaceMacros($t, [
'$submit' => DI::l10n()->t('Save Settings'),
'$good_crawlers' => ['good_crawlers', DI::l10n()->t('Allow "good" crawlers'), DI::config()->get('blockbot', 'good_crawlers'), "Don't block fediverse crawlers, relay servers and other bots with good purposes."],
'$block_gab' => ['block_gab', DI::l10n()->t('Block GabSocial'), DI::config()->get('blockbot', 'block_gab'), 'Block the software GabSocial. This will block every access for that software. You can block dedicated gab instances in the blocklist settings in the admin section.'],
'$training' => ['training', DI::l10n()->t('Training mode'), DI::config()->get('blockbot', 'training'), "Activates the training mode. This is only meant for developing purposes. Don't activate this on a production machine. This can cut communication with some systems."],
]);
}
function blockbot_addon_admin_post(&$a) {
DI::config()->set('blockbot', 'good_crawlers', $_POST['good_crawlers'] ?? false);
DI::config()->set('blockbot', 'block_gab', $_POST['block_gab'] ?? false);
DI::config()->set('blockbot', 'training', $_POST['training'] ?? false);
}
function blockbot_init_1(App $a) {
2019-05-03 12:25:13 +02:00
if (empty($_SERVER['HTTP_USER_AGENT'])) {
return;
}
$logdata = ['agent' => $_SERVER['HTTP_USER_AGENT'], 'uri' => $_SERVER['REQUEST_URI']];
2019-07-28 10:13:53 +02:00
// List of "good" crawlers
$good_agents = ['fediverse.space crawler', 'fediverse.network crawler', 'Active_Pods_CheckBot_3.0',
2020-09-06 09:30:59 +02:00
'Social-Relay/', 'Test Certificate Info', 'Uptimebot/', 'GNUSocialBot', 'UptimeRobot/',
'PTST/'];
2019-07-28 10:13:53 +02:00
2019-05-30 06:45:20 +02:00
// List of known crawlers.
2019-06-23 11:13:59 +02:00
$agents = ['SemrushBot', 's~feedly-nikon3', 'Qwantify/Bleriot/', 'ltx71', 'Sogou web spider/',
'Diffbot/', 'Twitterbot/', 'YisouSpider', 'evc-batch/', 'LivelapBot/', 'TrendsmapResolver/',
2019-05-29 20:51:07 +02:00
'PaperLiBot/', 'Nuzzel', 'um-LN/', 'Google Favicon', 'Datanyze', 'BLEXBot/', '360Spider',
'adscanner/', 'HeadlessChrome', 'wpif', 'startmebot/', 'Googlebot/', 'Applebot/',
2019-05-29 21:13:15 +02:00
'facebookexternalhit/', 'GoogleImageProxy', 'bingbot/', 'heritrix/', 'ldspider',
2019-05-31 17:37:33 +02:00
'AwarioRssBot/', 'Zabbix', 'TweetmemeBot/', 'dcrawl/', 'PhantomJS/', 'Googlebot-Image/',
2019-06-01 06:51:01 +02:00
'CrowdTanglebot/', 'Mediapartners-Google', 'Baiduspider/', 'datagnionbot',
2019-06-06 22:31:16 +02:00
'MegaIndex.ru/', 'SMUrlExpander', 'Hatena-Favicon/', 'Wappalyzer', 'FlipboardProxy/',
'NetcraftSurveyAgent/', 'Dataprovider.com', 'SMTBot/', 'Nimbostratus-Bot/',
2019-06-10 16:33:42 +02:00
'DuckDuckGo-Favicons-Bot/', 'IndieWebCards/', 'proximic', 'netEstate NE Crawler',
2019-06-17 16:07:37 +02:00
'AhrefsBot/', 'YandexBot/', 'Exabot/', 'Mediumbot-MetaTagFetcher/', 'WhatsApp/',
'TelegramBot', 'SurdotlyBot/', 'BingPreview/', 'SabsimBot/', 'CCBot/', 'WbSrch/',
2019-06-20 07:31:53 +02:00
'DuckDuckBot-Https/', 'HTTP Banner Detection', 'YandexImages/', 'archive.org_bot',
'ArchiveTeam ArchiveBot/', 'yacybot', 'https://developers.google.com/+/web/snippet/',
'Scrapy/', 'github-camo', 'MJ12bot/', 'DotBot/', 'Pinterestbot/', 'Jooblebot/',
2019-07-10 04:16:57 +02:00
'Cliqzbot/', 'YaK/', 'Mediatoolkitbot', 'Snacktory', 'FunWebProducts', 'oBot/',
2019-07-29 17:48:51 +02:00
'7Siters/', 'KOCMOHABT', 'Google-SearchByImage', 'FemtosearchBot/',
2020-07-05 16:36:17 +02:00
'HubSpot Crawler', 'DomainStatsBot/', 'Re-re Studio', 'AwarioSmartBot/',
2020-07-07 20:51:42 +02:00
'SummalyBot/', 'DNSResearchBot/', 'PetalBot;', 'Nmap Scripting Engine;',
2020-07-09 17:08:41 +02:00
'Google-Apps-Script; beanserver;', 'woorankreview/', 'Seekport Crawler;', 'AHC/',
2020-07-11 11:41:20 +02:00
'SkypeUriPreview Preview/', 'Semanticbot/', 'Embed PHP library', 'XoviOnpageCrawler;',
'GetHPinfo.com-Bot/', 'BoardReader Favicon Fetcher'];
if (!DI::config()->get('blockbot', 'good_crawlers')) {
2019-07-28 10:13:53 +02:00
$agents = array_merge($agents, $good_agents);
2020-01-18 23:28:14 +01:00
} else {
foreach ($good_agents as $good_agent) {
if (stristr($_SERVER['HTTP_USER_AGENT'], $good_agent)) {
return;
}
}
}
2019-07-28 10:13:53 +02:00
if (DI::config()->get('blockbot', 'block_gab')) {
$agents[] = 'GabSocial/';
}
2019-05-29 20:51:07 +02:00
foreach ($agents as $agent) {
if (stristr($_SERVER['HTTP_USER_AGENT'], $agent)) {
System::httpExit(403, 'Bots are not allowed');
}
}
2019-05-30 06:45:20 +02:00
// This switch here is only meant for developers who want to add more bots to the list above, it is not safe for production.
if (!DI::config()->get('blockbot', 'training')) {
2019-05-29 20:51:07 +02:00
return;
}
$crawlerDetect = new CrawlerDetect();
if (!$crawlerDetect->isCrawler()) {
logger::debug('Good user agent detected', $logdata);
return;
}
// List of false positives' strings of known "good" agents.
2019-07-28 10:13:53 +02:00
$agents = ['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 12:32:01 +02:00
'WordPress/', 'http.rb/', 'Apache-HttpClient/', 'WordPress.com;', 'Pleroma',
'Dispatch/', 'Ruby', 'Java/', 'libwww-perl/', 'Mastodon/', 'FeedlyApp/',
2020-07-08 21:24:00 +02:00
'lua-resty-http/', 'Tiny Tiny RSS/', 'Wget/', 'PostmanRuntime/',
'W3C_Validator/', 'NetNewsWire', 'FeedValidator/', 'theoldreader.com'];
2019-07-28 10:13:53 +02:00
if (DI::config()->get('blockbot', 'good_crawlers')) {
2019-07-28 10:13:53 +02:00
$agents = array_merge($agents, $good_agents);
}
foreach ($agents as $agent) {
if (stristr($_SERVER['HTTP_USER_AGENT'], $agent)) {
logger::notice('False positive', $logdata);
return;
}
}
logger::info('Blocked bot', $logdata);
2019-05-29 20:51:07 +02:00
System::httpExit(403, 'Bots are not allowed');
2019-04-20 14:15:45 +02:00
}