Merge pull request #2357 / limiter -> botdetection

The monolithic implementation of the limiter was divided into methods and
implemented in the Python package searx.botdetection.  Detailed documentation on
the methods has been added.

The methods are divided into two groups:

1. Probe HTTP headers

- Method http_accept
- Method http_accept_encoding
- Method http_accept_language
- Method http_connection
- Method http_user_agent

2. Rate limit:

- Method ip_limit
- Method link_token (new)

The (reduced) implementation of the limiter is now in the module
searx.botdetection.limiter.  The first group was transferred unchanged to this
module.  The ip_limit contains the sliding windows implemented by the limiter so
far.

This merge also fixes some long outstandig issue:

- limiter does not evaluate the Accept-Language correct [1]
- limiter needs a IPv6 prefix to block networks instead of IPs [2]

Without additional configuration the limiter works as before (apart from the
bugfixes).  For the commissioning of additional methods (link_toke), a
configuration must be made in an additional configuration file.  Without this
configuration, the limiter runs as before (zero configuration).

The ip_limit Method implements the sliding windows of the vanilla limiter,
additionally the link_token method can be used in this method.  The link_token
method can be used to investigate whether a request is suspicious. To activate
the link_token method in the ip_limit method add the following to your
/etc/searxng/limiter.toml::

    [botdetection.ip_limit]
    link_token = true


[1] https://github.com/searxng/searxng/issues/2455
[2] https://github.com/searxng/searxng/issues/2477
This commit is contained in:
Markus Heiser 2023-06-03 06:00:15 +02:00 committed by GitHub
commit 80aaef6c95
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
22 changed files with 1273 additions and 138 deletions

View file

@ -235,7 +235,7 @@ Global Settings
``limiter`` :
Rate limit the number of request on the instance, block some bots. The
:ref:`limiter plugin` requires a :ref:`settings redis` database.
:ref:`limiter src` requires a :ref:`settings redis` database.
.. _image_proxy:

View file

@ -0,0 +1,45 @@
.. _botdetection:
=============
Bot Detection
=============
.. contents:: Contents
:depth: 2
:local:
:backlinks: entry
.. automodule:: searx.botdetection
:members:
.. automodule:: searx.botdetection.limiter
:members:
Rate limit
==========
.. automodule:: searx.botdetection.ip_limit
:members:
.. automodule:: searx.botdetection.link_token
:members:
Probe HTTP headers
==================
.. automodule:: searx.botdetection.http_accept
:members:
.. automodule:: searx.botdetection.http_accept_encoding
:members:
.. automodule:: searx.botdetection.http_accept_language
:members:
.. automodule:: searx.botdetection.http_connection
:members:
.. automodule:: searx.botdetection.http_user_agent
:members:

View file

@ -1,13 +0,0 @@
.. _limiter plugin:
==============
Limiter Plugin
==============
.. sidebar:: info
The :ref:`limiter plugin` requires a :ref:`Redis <settings redis>` database.
.. automodule:: searx.plugins.limiter
:members:

View file

@ -16,3 +16,4 @@ redis==4.5.5
markdown-it-py==2.2.0
typing_extensions==4.6.3
fasttext-predict==0.9.2.1
pytomlpp==1.0.13

View file

@ -0,0 +1,27 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
""".. _botdetection src:
X-Forwarded-For
===============
.. attention::
A correct setup of the HTTP request headers ``X-Forwarded-For`` and
``X-Real-IP`` is essential to be able to assign a request to an IP correctly:
- `NGINX RequestHeader`_
- `Apache RequestHeader`_
.. _NGINX RequestHeader:
https://docs.searxng.org/admin/installation-nginx.html#nginx-s-searxng-site
.. _Apache RequestHeader:
https://docs.searxng.org/admin/installation-apache.html#apache-s-searxng-site
.. autofunction:: searx.botdetection.get_real_ip
"""
from ._helpers import dump_request
from ._helpers import get_real_ip
from ._helpers import too_many_requests

View file

@ -0,0 +1,121 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
# pylint: disable=missing-module-docstring, invalid-name
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
IPv6Address,
ip_address,
ip_network,
)
import flask
import werkzeug
from searx.tools import config
from searx import logger
logger = logger.getChild('botdetection')
def dump_request(request: flask.Request):
return (
request.path
+ " || X-Forwarded-For: %s" % request.headers.get('X-Forwarded-For')
+ " || X-Real-IP: %s" % request.headers.get('X-Real-IP')
+ " || form: %s" % request.form
+ " || Accept: %s" % request.headers.get('Accept')
+ " || Accept-Language: %s" % request.headers.get('Accept-Language')
+ " || Accept-Encoding: %s" % request.headers.get('Accept-Encoding')
+ " || Content-Type: %s" % request.headers.get('Content-Type')
+ " || Content-Length: %s" % request.headers.get('Content-Length')
+ " || Connection: %s" % request.headers.get('Connection')
+ " || User-Agent: %s" % request.headers.get('User-Agent')
)
def too_many_requests(network: IPv4Network | IPv6Network, log_msg: str) -> werkzeug.Response | None:
"""Returns a HTTP 429 response object and writes a ERROR message to the
'botdetection' logger. This function is used in part by the filter methods
to return the default ``Too Many Requests`` response.
"""
logger.debug("BLOCK %s: %s", network.compressed, log_msg)
return flask.make_response(('Too Many Requests', 429))
def get_network(real_ip: str, cfg: config.Config) -> IPv4Network | IPv6Network:
"""Returns the (client) network of whether the real_ip is part of."""
ip = ip_address(real_ip)
if isinstance(ip, IPv6Address):
prefix = cfg['real_ip.ipv6_prefix']
else:
prefix = cfg['real_ip.ipv4_prefix']
network = ip_network(f"{real_ip}/{prefix}", strict=False)
# logger.debug("get_network(): %s", network.compressed)
return network
def get_real_ip(request: flask.Request) -> str:
"""Returns real IP of the request. Since not all proxies set all the HTTP
headers and incoming headers can be faked it may happen that the IP cannot
be determined correctly.
.. sidebar:: :py:obj:`flask.Request.remote_addr`
SearXNG uses Werkzeug's ProxyFix_ (with it default ``x_for=1``).
This function tries to get the remote IP in the order listed below,
additional some tests are done and if inconsistencies or errors are
detected, they are logged.
The remote IP of the request is taken from (first match):
- X-Forwarded-For_ header
- `X-real-IP header <https://github.com/searxng/searxng/issues/1237#issuecomment-1147564516>`__
- :py:obj:`flask.Request.remote_addr`
.. _ProxyFix:
https://werkzeug.palletsprojects.com/middleware/proxy_fix/
.. _X-Forwarded-For:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Forwarded-For
"""
forwarded_for = request.headers.get("X-Forwarded-For")
real_ip = request.headers.get('X-Real-IP')
remote_addr = request.remote_addr
# logger.debug(
# "X-Forwarded-For: %s || X-Real-IP: %s || request.remote_addr: %s", forwarded_for, real_ip, remote_addr
# )
if not forwarded_for:
logger.error("X-Forwarded-For header is not set!")
else:
from .limiter import get_cfg # pylint: disable=import-outside-toplevel, cyclic-import
forwarded_for = [x.strip() for x in forwarded_for.split(',')]
x_for: int = get_cfg()['real_ip.x_for']
forwarded_for = forwarded_for[-min(len(forwarded_for), x_for)]
if not real_ip:
logger.error("X-Real-IP header is not set!")
if forwarded_for and real_ip and forwarded_for != real_ip:
logger.warning("IP from X-Real-IP (%s) is not equal to IP from X-Forwarded-For (%s)", real_ip, forwarded_for)
if forwarded_for and remote_addr and forwarded_for != remote_addr:
logger.warning(
"IP from WSGI environment (%s) is not equal to IP from X-Forwarded-For (%s)", remote_addr, forwarded_for
)
if real_ip and remote_addr and real_ip != remote_addr:
logger.warning("IP from WSGI environment (%s) is not equal to IP from X-Real-IP (%s)", remote_addr, real_ip)
request_ip = forwarded_for or real_ip or remote_addr or '0.0.0.0'
# logger.debug("get_real_ip() -> %s", request_ip)
return request_ip

View file

@ -0,0 +1,39 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``http_accept``
----------------------
The ``http_accept`` method evaluates a request as the request of a bot if the
Accept_ header ..
- did not contain ``text/html``
.. _Accept:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Accept
"""
# pylint: disable=unused-argument
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from ._helpers import too_many_requests
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
if 'text/html' not in request.accept_mimetypes:
return too_many_requests(network, "HTTP header Accept did not contain text/html")
return None

View file

@ -0,0 +1,41 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``http_accept_encoding``
-------------------------------
The ``http_accept_encoding`` method evaluates a request as the request of a
bot if the Accept-Encoding_ header ..
- did not contain ``gzip`` AND ``deflate`` (if both values are missed)
- did not contain ``text/html``
.. _Accept-Encoding:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Accept-Encoding
"""
# pylint: disable=unused-argument
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from ._helpers import too_many_requests
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
accept_list = [l.strip() for l in request.headers.get('Accept-Encoding', '').split(',')]
if not ('gzip' in accept_list or 'deflate' in accept_list):
return too_many_requests(network, "HTTP header Accept-Encoding did not contain gzip nor deflate")
return None

View file

@ -0,0 +1,35 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``http_accept_language``
-------------------------------
The ``http_accept_language`` method evaluates a request as the request of a bot
if the Accept-Language_ header is unset.
.. _Accept-Language:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/User-Agent
"""
# pylint: disable=unused-argument
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from ._helpers import too_many_requests
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
if request.headers.get('Accept-Language', '').strip() == '':
return too_many_requests(network, "missing HTTP header Accept-Language")
return None

View file

@ -0,0 +1,37 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``http_connection``
--------------------------
The ``http_connection`` method evaluates a request as the request of a bot if
the Connection_ header is set to ``close``.
.. _Connection:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Connection
"""
# pylint: disable=unused-argument
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from ._helpers import too_many_requests
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
if request.headers.get('Connection', '').strip() == 'close':
return too_many_requests(network, "HTTP header 'Connection=close")
return None

View file

@ -0,0 +1,67 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``http_user_agent``
--------------------------
The ``http_user_agent`` method evaluates a request as the request of a bot if
the User-Agent_ header is unset or matches the regular expression
:py:obj:`USER_AGENT`.
.. _User-Agent:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/User-Agent
"""
# pylint: disable=unused-argument
from __future__ import annotations
import re
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from ._helpers import too_many_requests
USER_AGENT = (
r'('
+ r'unknown'
+ r'|[Cc][Uu][Rr][Ll]|[wW]get|Scrapy|splash|JavaFX|FeedFetcher|python-requests|Go-http-client|Java|Jakarta|okhttp'
+ r'|HttpClient|Jersey|Python|libwww-perl|Ruby|SynHttpClient|UniversalFeedParser|Googlebot|GoogleImageProxy'
+ r'|bingbot|Baiduspider|yacybot|YandexMobileBot|YandexBot|Yahoo! Slurp|MJ12bot|AhrefsBot|archive.org_bot|msnbot'
+ r'|MJ12bot|SeznamBot|linkdexbot|Netvibes|SMTBot|zgrab|James BOT|Sogou|Abonti|Pixray|Spinn3r|SemrushBot|Exabot'
+ r'|ZmEu|BLEXBot|bitlybot'
# unmaintained Farside instances
+ r'|'
+ re.escape(r'Mozilla/5.0 (compatible; Farside/0.1.0; +https://farside.link)')
# other bots and client to block
+ '|.*PetalBot.*'
+ r')'
)
"""Regular expression that matches to User-Agent_ from known *bots*"""
_regexp = None
def regexp_user_agent():
global _regexp # pylint: disable=global-statement
if not _regexp:
_regexp = re.compile(USER_AGENT)
return _regexp
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
user_agent = request.headers.get('User-Agent', 'unknown')
if regexp_user_agent().match(user_agent):
return too_many_requests(network, f"bot detected, HTTP header User-Agent: {user_agent}")
return None

View file

@ -0,0 +1,146 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
""".. _botdetection.ip_limit:
Method ``ip_limit``
-------------------
The ``ip_limit`` method counts request from an IP in *sliding windows*. If
there are to many requests in a sliding window, the request is evaluated as a
bot request. This method requires a redis DB and needs a HTTP X-Forwarded-For_
header. To take privacy only the hash value of an IP is stored in the redis DB
and at least for a maximum of 10 minutes.
The :py:obj:`.link_token` method can be used to investigate whether a request is
*suspicious*. To activate the :py:obj:`.link_token` method in the
:py:obj:`.ip_limit` method add the following to your
``/etc/searxng/limiter.toml``:
.. code:: toml
[botdetection.ip_limit]
link_token = true
If the :py:obj:`.link_token` method is activated and a request is *suspicious*
the request rates are reduced:
- :py:obj:`BURST_MAX` -> :py:obj:`BURST_MAX_SUSPICIOUS`
- :py:obj:`LONG_MAX` -> :py:obj:`LONG_MAX_SUSPICIOUS`
To intercept bots that get their IPs from a range of IPs, there is a
:py:obj:`SUSPICIOUS_IP_WINDOW`. In this window the suspicious IPs are stored
for a longer time. IPs stored in this sliding window have a maximum of
:py:obj:`SUSPICIOUS_IP_MAX` accesses before they are blocked. As soon as the IP
makes a request that is not suspicious, the sliding window for this IP is
droped.
.. _X-Forwarded-For:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Forwarded-For
"""
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import flask
import werkzeug
from searx.tools import config
from searx import redisdb
from searx import logger
from searx.redislib import incr_sliding_window, drop_counter
from . import link_token
from ._helpers import too_many_requests
logger = logger.getChild('botdetection.ip_limit')
BURST_WINDOW = 20
"""Time (sec) before sliding window for *burst* requests expires."""
BURST_MAX = 15
"""Maximum requests from one IP in the :py:obj:`BURST_WINDOW`"""
BURST_MAX_SUSPICIOUS = 2
"""Maximum of suspicious requests from one IP in the :py:obj:`BURST_WINDOW`"""
LONG_WINDOW = 600
"""Time (sec) before the longer sliding window expires."""
LONG_MAX = 150
"""Maximum requests from one IP in the :py:obj:`LONG_WINDOW`"""
LONG_MAX_SUSPICIOUS = 10
"""Maximum suspicious requests from one IP in the :py:obj:`LONG_WINDOW`"""
API_WONDOW = 3600
"""Time (sec) before sliding window for API requests (format != html) expires."""
API_MAX = 4
"""Maximum requests from one IP in the :py:obj:`API_WONDOW`"""
SUSPICIOUS_IP_WINDOW = 3600 * 24 * 30
"""Time (sec) before sliding window for one suspicious IP expires."""
SUSPICIOUS_IP_MAX = 3
"""Maximum requests from one suspicious IP in the :py:obj:`SUSPICIOUS_IP_WINDOW`."""
def filter_request(
network: IPv4Network | IPv6Network,
request: flask.Request,
cfg: config.Config,
) -> werkzeug.Response | None:
# pylint: disable=too-many-return-statements
redis_client = redisdb.client()
if network.is_link_local and not cfg['botdetection.ip_limit.filter_link_local']:
logger.debug("network %s is link-local -> not monitored by ip_limit method", network.compressed)
return None
if request.args.get('format', 'html') != 'html':
c = incr_sliding_window(redis_client, 'ip_limit.API_WONDOW:' + network.compressed, API_WONDOW)
if c > API_MAX:
return too_many_requests(network, "too many request in API_WINDOW")
if cfg['botdetection.ip_limit.link_token']:
suspicious = link_token.is_suspicious(network, request, True)
if not suspicious:
# this IP is no longer suspicious: release ip again / delete the counter of this IP
drop_counter(redis_client, 'ip_limit.SUSPICIOUS_IP_WINDOW' + network.compressed)
return None
# this IP is suspicious: count requests from this IP
c = incr_sliding_window(
redis_client, 'ip_limit.SUSPICIOUS_IP_WINDOW' + network.compressed, SUSPICIOUS_IP_WINDOW
)
if c > SUSPICIOUS_IP_MAX:
logger.error("BLOCK: too many request from %s in SUSPICIOUS_IP_WINDOW (redirect to /)", network)
return flask.redirect(flask.url_for('index'), code=302)
c = incr_sliding_window(redis_client, 'ip_limit.BURST_WINDOW' + network.compressed, BURST_WINDOW)
if c > BURST_MAX_SUSPICIOUS:
return too_many_requests(network, "too many request in BURST_WINDOW (BURST_MAX_SUSPICIOUS)")
c = incr_sliding_window(redis_client, 'ip_limit.LONG_WINDOW' + network.compressed, LONG_WINDOW)
if c > LONG_MAX_SUSPICIOUS:
return too_many_requests(network, "too many request in LONG_WINDOW (LONG_MAX_SUSPICIOUS)")
return None
# vanilla limiter without extensions counts BURST_MAX and LONG_MAX
c = incr_sliding_window(redis_client, 'ip_limit.BURST_WINDOW' + network.compressed, BURST_WINDOW)
if c > BURST_MAX:
return too_many_requests(network, "too many request in BURST_WINDOW (BURST_MAX)")
c = incr_sliding_window(redis_client, 'ip_limit.LONG_WINDOW' + network.compressed, LONG_WINDOW)
if c > LONG_MAX:
return too_many_requests(network, "too many request in LONG_WINDOW (LONG_MAX)")
return None

View file

@ -0,0 +1,118 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
""".. _limiter src:
Limiter
=======
.. sidebar:: info
The limiter requires a :ref:`Redis <settings redis>` database.
Bot protection / IP rate limitation. The intention of rate limitation is to
limit suspicious requests from an IP. The motivation behind this is the fact
that SearXNG passes through requests from bots and is thus classified as a bot
itself. As a result, the SearXNG engine then receives a CAPTCHA or is blocked
by the search engine (the origin) in some other way.
To avoid blocking, the requests from bots to SearXNG must also be blocked, this
is the task of the limiter. To perform this task, the limiter uses the methods
from the :py:obj:`searx.botdetection`.
To enable the limiter activate:
.. code:: yaml
server:
...
limiter: true # rate limit the number of request on the instance, block some bots
and set the redis-url connection. Check the value, it depends on your redis DB
(see :ref:`settings redis`), by example:
.. code:: yaml
redis:
url: unix:///usr/local/searxng-redis/run/redis.sock?db=0
"""
from __future__ import annotations
from pathlib import Path
import flask
import werkzeug
from searx.tools import config
from searx import logger
from . import (
http_accept,
http_accept_encoding,
http_accept_language,
http_connection,
http_user_agent,
ip_limit,
)
from ._helpers import (
get_network,
get_real_ip,
dump_request,
)
logger = logger.getChild('botdetection.limiter')
CFG: config.Config = None # type: ignore
LIMITER_CFG_SCHEMA = Path(__file__).parent / "limiter.toml"
"""Base configuration (schema) of the botdetection."""
LIMITER_CFG = Path('/etc/searxng/limiter.toml')
"""Lokal Limiter configuration."""
CFG_DEPRECATED = {
# "dummy.old.foo": "config 'dummy.old.foo' exists only for tests. Don't use it in your real project config."
}
def get_cfg() -> config.Config:
global CFG # pylint: disable=global-statement
if CFG is None:
CFG = config.Config.from_toml(LIMITER_CFG_SCHEMA, LIMITER_CFG, CFG_DEPRECATED)
return CFG
def filter_request(request: flask.Request) -> werkzeug.Response | None:
cfg = get_cfg()
real_ip = get_real_ip(request)
network = get_network(real_ip, cfg)
if network.is_link_local:
return None
if request.path == '/healthz':
return None
for func in [
http_user_agent,
]:
val = func.filter_request(network, request, cfg)
if val is not None:
return val
if request.path == '/search':
for func in [
http_accept,
http_accept_encoding,
http_accept_language,
http_connection,
http_user_agent,
ip_limit,
]:
val = func.filter_request(network, request, cfg)
if val is not None:
return val
logger.debug(f"OK {network}: %s", dump_request(flask.request))
return None

View file

@ -0,0 +1,22 @@
[real_ip]
# Number of values to trust for X-Forwarded-For.
x_for = 1
# The prefix defines the number of leading bits in an address that are compared
# to determine whether or not an address is part of a (client) network.
ipv4_prefix = 32
ipv6_prefix = 48
[botdetection.ip_limit]
# To get unlimited access in a local network, by default link-lokal addresses
# (networks) are not monitored by the ip_limit
filter_link_local = false
# acrivate link_token method in the ip_limit method
link_token = false

View file

@ -0,0 +1,156 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Method ``link_token``
---------------------
The ``link_token`` method evaluates a request as :py:obj:`suspicious
<is_suspicious>` if the URL ``/client<token>.css`` is not requested by the
client. By adding a random component (the token) in the URL, a bot can not send
a ping by request a static URL.
.. note::
This method requires a redis DB and needs a HTTP X-Forwarded-For_ header.
To get in use of this method a flask URL route needs to be added:
.. code:: python
@app.route('/client<token>.css', methods=['GET', 'POST'])
def client_token(token=None):
link_token.ping(request, token)
return Response('', mimetype='text/css')
And in the HTML template from flask a stylesheet link is needed (the value of
``link_token`` comes from :py:obj:`get_token`):
.. code:: html
<link rel="stylesheet"
href="{{ url_for('client_token', token=link_token) }}"
type="text/css" />
.. _X-Forwarded-For:
https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Forwarded-For
"""
from __future__ import annotations
from ipaddress import (
IPv4Network,
IPv6Network,
)
import string
import random
import flask
from searx import logger
from searx import redisdb
from searx.redislib import secret_hash
from ._helpers import (
get_network,
get_real_ip,
)
TOKEN_LIVE_TIME = 600
"""Livetime (sec) of limiter's CSS token."""
PING_LIVE_TIME = 3600
"""Livetime (sec) of the ping-key from a client (request)"""
PING_KEY = 'SearXNG_limiter.ping'
"""Prefix of all ping-keys generated by :py:obj:`get_ping_key`"""
TOKEN_KEY = 'SearXNG_limiter.token'
"""Key for which the current token is stored in the DB"""
logger = logger.getChild('botdetection.link_token')
def is_suspicious(network: IPv4Network | IPv6Network, request: flask.Request, renew: bool = False):
"""Checks whether a valid ping is exists for this (client) network, if not
this request is rated as *suspicious*. If a valid ping exists and argument
``renew`` is ``True`` the expire time of this ping is reset to
:py:obj:`PING_LIVE_TIME`.
"""
redis_client = redisdb.client()
if not redis_client:
return False
ping_key = get_ping_key(network, request)
if not redis_client.get(ping_key):
logger.warning("missing ping (IP: %s) / request: %s", network.compressed, ping_key)
return True
if renew:
redis_client.set(ping_key, 1, ex=PING_LIVE_TIME)
logger.debug("found ping for (client) network %s -> %s", network.compressed, ping_key)
return False
def ping(request: flask.Request, token: str):
"""This function is called by a request to URL ``/client<token>.css``. If
``token`` is valid a :py:obj:`PING_KEY` for the client is stored in the DB.
The expire time of this ping-key is :py:obj:`PING_LIVE_TIME`.
"""
from . import limiter # pylint: disable=import-outside-toplevel, cyclic-import
redis_client = redisdb.client()
if not redis_client:
return
if not token_is_valid(token):
return
cfg = limiter.get_cfg()
real_ip = get_real_ip(request)
network = get_network(real_ip, cfg)
ping_key = get_ping_key(network, request)
logger.debug("store ping_key for (client) network %s (IP %s) -> %s", network.compressed, real_ip, ping_key)
redis_client.set(ping_key, 1, ex=PING_LIVE_TIME)
def get_ping_key(network: IPv4Network | IPv6Network, request: flask.Request) -> str:
"""Generates a hashed key that fits (more or less) to a *WEB-browser
session* in a network."""
return (
PING_KEY
+ "["
+ secret_hash(
network.compressed + request.headers.get('Accept-Language', '') + request.headers.get('User-Agent', '')
)
+ "]"
)
def token_is_valid(token) -> bool:
valid = token == get_token()
logger.debug("token is valid --> %s", valid)
return valid
def get_token() -> str:
"""Returns current token. If there is no currently active token a new token
is generated randomly and stored in the redis DB.
- :py:obj:`TOKEN_LIVE_TIME`
- :py:obj:`TOKEN_KEY`
"""
redis_client = redisdb.client()
if not redis_client:
# This function is also called when limiter is inactive / no redis DB
# (see render function in webapp.py)
return '12345678'
token = redis_client.get(TOKEN_KEY)
if token:
token = token.decode('UTF-8')
else:
token = ''.join(random.choice(string.ascii_lowercase + string.digits) for _ in range(16))
redis_client.set(TOKEN_KEY, token, ex=TOKEN_LIVE_TIME)
return token

View file

@ -1,119 +1,32 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
# pyright: basic
"""Some bot protection / rate limitation
"""see :ref:`limiter src`"""
To monitor rate limits and protect privacy the IP addresses are getting stored
with a hash so the limiter plugin knows who to block. A redis database is
needed to store the hash values.
Enable the plugin in ``settings.yml``:
- ``server.limiter: true``
- ``redis.url: ...`` check the value, see :ref:`settings redis`
"""
import re
from flask import request
import flask
from searx import redisdb
from searx.plugins import logger
from searx.redislib import incr_sliding_window
from searx.botdetection import limiter
name = "Request limiter"
description = "Limit the number of request"
default_on = False
preference_section = 'service'
logger = logger.getChild('limiter')
block_user_agent = re.compile(
r'('
+ r'unknown'
+ r'|[Cc][Uu][Rr][Ll]|[wW]get|Scrapy|splash|JavaFX|FeedFetcher|python-requests|Go-http-client|Java|Jakarta|okhttp'
+ r'|HttpClient|Jersey|Python|libwww-perl|Ruby|SynHttpClient|UniversalFeedParser|Googlebot|GoogleImageProxy'
+ r'|bingbot|Baiduspider|yacybot|YandexMobileBot|YandexBot|Yahoo! Slurp|MJ12bot|AhrefsBot|archive.org_bot|msnbot'
+ r'|MJ12bot|SeznamBot|linkdexbot|Netvibes|SMTBot|zgrab|James BOT|Sogou|Abonti|Pixray|Spinn3r|SemrushBot|Exabot'
+ r'|ZmEu|BLEXBot|bitlybot'
# unmaintained Farside instances
+ r'|'
+ re.escape(r'Mozilla/5.0 (compatible; Farside/0.1.0; +https://farside.link)')
+ '|.*PetalBot.*'
+ r')'
)
def is_accepted_request() -> bool:
# pylint: disable=too-many-return-statements
redis_client = redisdb.client()
user_agent = request.headers.get('User-Agent', 'unknown')
x_forwarded_for = request.headers.get('X-Forwarded-For', '')
if request.path == '/healthz':
return True
if block_user_agent.match(user_agent):
logger.debug("BLOCK %s: %s --> detected User-Agent: %s" % (x_forwarded_for, request.path, user_agent))
return False
if request.path == '/search':
c_burst = incr_sliding_window(redis_client, 'IP limit, burst' + x_forwarded_for, 20)
c_10min = incr_sliding_window(redis_client, 'IP limit, 10 minutes' + x_forwarded_for, 600)
if c_burst > 15 or c_10min > 150:
logger.debug("BLOCK %s: to many request", x_forwarded_for)
return False
if len(request.headers.get('Accept-Language', '').strip()) == '':
logger.debug("BLOCK %s: missing Accept-Language", x_forwarded_for)
return False
if request.headers.get('Connection') == 'close':
logger.debug("BLOCK %s: got Connection=close", x_forwarded_for)
return False
accept_encoding_list = [l.strip() for l in request.headers.get('Accept-Encoding', '').split(',')]
if 'gzip' not in accept_encoding_list and 'deflate' not in accept_encoding_list:
logger.debug("BLOCK %s: suspicious Accept-Encoding", x_forwarded_for)
return False
if 'text/html' not in request.accept_mimetypes:
logger.debug("BLOCK %s: Accept-Encoding misses text/html", x_forwarded_for)
return False
if request.args.get('format', 'html') != 'html':
c = incr_sliding_window(redis_client, 'API limit' + x_forwarded_for, 3600)
if c > 4:
logger.debug("BLOCK %s: API limit exceeded", x_forwarded_for)
return False
logger.debug(
"OK %s: '%s'" % (x_forwarded_for, request.path)
+ " || form: %s" % request.form
+ " || Accept: %s" % request.headers.get('Accept', '')
+ " || Accept-Language: %s" % request.headers.get('Accept-Language', '')
+ " || Accept-Encoding: %s" % request.headers.get('Accept-Encoding', '')
+ " || Content-Type: %s" % request.headers.get('Content-Type', '')
+ " || Content-Length: %s" % request.headers.get('Content-Length', '')
+ " || Connection: %s" % request.headers.get('Connection', '')
+ " || User-Agent: %s" % user_agent
)
return True
def pre_request():
if not is_accepted_request():
return 'Too Many Requests', 429
return None
"""See :ref:`flask.Flask.before_request`"""
return limiter.filter_request(flask.request)
def init(app, settings):
def init(app: flask.Flask, settings) -> bool:
if not settings['server']['limiter']:
return False
if not redisdb.client():
logger.error("The limiter requires Redis") # pylint: disable=undefined-variable
logger.error("The limiter requires Redis")
return False
app.before_request(pre_request)
return True

View file

@ -1,21 +1,11 @@
'''
searx is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
# pylint: disable=missing-module-docstring,invalid-name
searx is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with searx. If not, see < http://www.gnu.org/licenses/ >.
(C) 2015 by Adam Tauber, <asciimoo@gmail.com>
'''
from flask_babel import gettext
import re
from flask_babel import gettext
from searx.botdetection._helpers import get_real_ip
name = gettext('Self Information')
description = gettext('Displays your IP if the query is "ip" and your user agent if the query contains "user agent".')
@ -28,18 +18,11 @@ query_examples = ''
p = re.compile('.*user[ -]agent.*', re.IGNORECASE)
# attach callback to the post search hook
# request: flask request object
# ctx: the whole local context of the pre search hook
def post_search(request, search):
if search.search_query.pageno > 1:
return True
if search.search_query.query == 'ip':
x_forwarded_for = request.headers.getlist("X-Forwarded-For")
if x_forwarded_for:
ip = x_forwarded_for[0]
else:
ip = request.remote_addr
ip = get_real_ip(request)
search.result_container.answers['ip'] = {'answer': ip}
elif p.match(search.search_query.query):
ua = request.user_agent

View file

@ -17,6 +17,9 @@
{% else %}
<link rel="stylesheet" href="{{ url_for('static', filename='css/searxng.min.css') }}" type="text/css" media="screen" />
{% endif %}
{% if get_setting('server.limiter') %}
<link rel="stylesheet" href="{{ url_for('client_token', token=link_token) }}" type="text/css" />
{% endif %}
{% block styles %}{% endblock %}
<!--[if gte IE 9]>-->
<script src="{{ url_for('static', filename='js/searxng.head.min.js') }}" client_settings="{{ client_settings }}"></script>

8
searx/tools/__init__.py Normal file
View file

@ -0,0 +1,8 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
""".. _tools src:
A collection of *utilities* used by SearXNG, but without SearXNG specific
peculiarities.
"""

376
searx/tools/config.py Normal file
View file

@ -0,0 +1,376 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""Configuration class :py:class:`Config` with deep-update, schema validation
and deprecated names.
The :py:class:`Config` class implements a configuration that is based on
structured dictionaries. The configuration schema is defined in a dictionary
structure and the configuration data is given in a dictionary structure.
"""
from __future__ import annotations
import copy
import typing
import logging
import pathlib
import pytomlpp as toml
__all__ = ['Config', 'UNSET', 'SchemaIssue']
log = logging.getLogger(__name__)
class FALSE:
"""Class of ``False`` singelton"""
# pylint: disable=multiple-statements
def __init__(self, msg):
self.msg = msg
def __bool__(self):
return False
def __str__(self):
return self.msg
__repr__ = __str__
UNSET = FALSE('<UNSET>')
class SchemaIssue(ValueError):
"""Exception to store and/or raise a message from a schema issue."""
def __init__(self, level: typing.Literal['warn', 'invalid'], msg: str):
self.level = level
super().__init__(msg)
def __str__(self):
return f"[cfg schema {self.level}] {self.args[0]}"
class Config:
"""Base class used for configuration"""
UNSET = UNSET
@classmethod
def from_toml(cls, schema_file: pathlib.Path, cfg_file: pathlib.Path, deprecated: dict) -> Config:
# init schema
log.debug("load schema file: %s", schema_file)
cfg = cls(cfg_schema=toml.load(schema_file), deprecated=deprecated)
if not cfg_file.exists():
log.warning("missing config file: %s", cfg_file)
return cfg
# load configuration
log.debug("load config file: %s", cfg_file)
try:
upd_cfg = toml.load(cfg_file)
except toml.DecodeError as exc:
msg = str(exc).replace('\t', '').replace('\n', ' ')
log.error("%s: %s", cfg_file, msg)
raise
is_valid, issue_list = cfg.validate(upd_cfg)
for msg in issue_list:
log.error(str(msg))
if not is_valid:
raise TypeError(f"schema of {cfg_file} is invalid!")
cfg.update(upd_cfg)
return cfg
def __init__(self, cfg_schema: typing.Dict, deprecated: typing.Dict[str, str]):
"""Construtor of class Config.
:param cfg_schema: Schema of the configuration
:param deprecated: dictionary that maps deprecated configuration names to a messages
These values are needed for validation, see :py:obj:`validate`.
"""
self.cfg_schema = cfg_schema
self.deprecated = deprecated
self.cfg = copy.deepcopy(cfg_schema)
def __getitem__(self, key: str):
return self.get(key)
def validate(self, cfg: dict):
"""Validation of dictionary ``cfg`` on :py:obj:`Config.SCHEMA`.
Validation is done by :py:obj:`validate`."""
return validate(self.cfg_schema, cfg, self.deprecated)
def update(self, upd_cfg: dict):
"""Update this configuration by ``upd_cfg``."""
dict_deepupdate(self.cfg, upd_cfg)
def default(self, name: str):
"""Returns default value of field ``name`` in ``self.cfg_schema``."""
return value(name, self.cfg_schema)
def get(self, name: str, default=UNSET, replace=True):
"""Returns the value to which ``name`` points in the configuration.
If there is no such ``name`` in the config and the ``default`` is
:py:obj:`UNSET`, a :py:obj:`KeyError` is raised.
"""
parent = self._get_parent_dict(name)
val = parent.get(name.split('.')[-1], UNSET)
if val is UNSET:
if default is UNSET:
raise KeyError(name)
val = default
if replace and isinstance(val, str):
val = val % self
return val
def set(self, name: str, val):
"""Set the value to which ``name`` points in the configuration.
If there is no such ``name`` in the config, a :py:obj:`KeyError` is
raised.
"""
parent = self._get_parent_dict(name)
parent[name.split('.')[-1]] = val
def _get_parent_dict(self, name):
parent_name = '.'.join(name.split('.')[:-1])
if parent_name:
parent = value(parent_name, self.cfg)
else:
parent = self.cfg
if (parent is UNSET) or (not isinstance(parent, dict)):
raise KeyError(parent_name)
return parent
def path(self, name: str, default=UNSET):
"""Get a :py:class:`pathlib.Path` object from a config string."""
val = self.get(name, default)
if val is UNSET:
if default is UNSET:
raise KeyError(name)
return default
return pathlib.Path(str(val))
def pyobj(self, name, default=UNSET):
"""Get python object refered by full qualiffied name (FQN) in the config
string."""
fqn = self.get(name, default)
if fqn is UNSET:
if default is UNSET:
raise KeyError(name)
return default
(modulename, name) = str(fqn).rsplit('.', 1)
m = __import__(modulename, {}, {}, [name], 0)
return getattr(m, name)
# working with dictionaries
def value(name: str, data_dict: dict):
"""Returns the value to which ``name`` points in the ``dat_dict``.
.. code: python
>>> data_dict = {
"foo": {"bar": 1 },
"bar": {"foo": 2 },
"foobar": [1, 2, 3],
}
>>> value('foobar', data_dict)
[1, 2, 3]
>>> value('foo.bar', data_dict)
1
>>> value('foo.bar.xxx', data_dict)
<UNSET>
"""
ret_val = data_dict
for part in name.split('.'):
if isinstance(ret_val, dict):
ret_val = ret_val.get(part, UNSET)
if ret_val is UNSET:
break
return ret_val
def validate(
schema_dict: typing.Dict, data_dict: typing.Dict, deprecated: typing.Dict[str, str]
) -> typing.Tuple[bool, list]:
"""Deep validation of dictionary in ``data_dict`` against dictionary in
``schema_dict``. Argument deprecated is a dictionary that maps deprecated
configuration names to a messages::
deprecated = {
"foo.bar" : "config 'foo.bar' is deprecated, use 'bar.foo'",
"..." : "..."
}
The function returns a python tuple ``(is_valid, issue_list)``:
``is_valid``:
A bool value indicating ``data_dict`` is valid or not.
``issue_list``:
A list of messages (:py:obj:`SchemaIssue`) from the validation::
[schema warn] data_dict: deprecated 'fontlib.foo': <DEPRECATED['foo.bar']>
[schema invalid] data_dict: key unknown 'fontlib.foo'
[schema invalid] data_dict: type mismatch 'fontlib.foo': expected ..., is ...
If ``schema_dict`` or ``data_dict`` is not a dictionary type a
:py:obj:`SchemaIssue` is raised.
"""
names = []
is_valid = True
issue_list = []
if not isinstance(schema_dict, dict):
raise SchemaIssue('invalid', "schema_dict is not a dict type")
if not isinstance(data_dict, dict):
raise SchemaIssue('invalid', f"data_dict issue{'.'.join(names)} is not a dict type")
is_valid, issue_list = _validate(names, issue_list, schema_dict, data_dict, deprecated)
return is_valid, issue_list
def _validate(
names: typing.List,
issue_list: typing.List,
schema_dict: typing.Dict,
data_dict: typing.Dict,
deprecated: typing.Dict[str, str],
) -> typing.Tuple[bool, typing.List]:
is_valid = True
for key, data_value in data_dict.items():
names.append(key)
name = '.'.join(names)
deprecated_msg = deprecated.get(name)
# print("XXX %s: key %s // data_value: %s" % (name, key, data_value))
if deprecated_msg:
issue_list.append(SchemaIssue('warn', f"data_dict '{name}': deprecated - {deprecated_msg}"))
schema_value = value(name, schema_dict)
# print("YYY %s: key %s // schema_value: %s" % (name, key, schema_value))
if schema_value is UNSET:
if not deprecated_msg:
issue_list.append(SchemaIssue('invalid', f"data_dict '{name}': key unknown in schema_dict"))
is_valid = False
elif type(schema_value) != type(data_value): # pylint: disable=unidiomatic-typecheck
issue_list.append(
SchemaIssue(
'invalid',
(f"data_dict: type mismatch '{name}':" f" expected {type(schema_value)}, is: {type(data_value)}"),
)
)
is_valid = False
elif isinstance(data_value, dict):
_valid, _ = _validate(names, issue_list, schema_dict, data_value, deprecated)
is_valid = is_valid and _valid
names.pop()
return is_valid, issue_list
def dict_deepupdate(base_dict: dict, upd_dict: dict, names=None):
"""Deep-update of dictionary in ``base_dict`` by dictionary in ``upd_dict``.
For each ``upd_key`` & ``upd_val`` pair in ``upd_dict``:
0. If types of ``base_dict[upd_key]`` and ``upd_val`` do not match raise a
:py:obj:`TypeError`.
1. If ``base_dict[upd_key]`` is a dict: recursively deep-update it by ``upd_val``.
2. If ``base_dict[upd_key]`` not exist: set ``base_dict[upd_key]`` from a
(deep-) copy of ``upd_val``.
3. If ``upd_val`` is a list, extend list in ``base_dict[upd_key]`` by the
list in ``upd_val``.
4. If ``upd_val`` is a set, update set in ``base_dict[upd_key]`` by set in
``upd_val``.
"""
# pylint: disable=too-many-branches
if not isinstance(base_dict, dict):
raise TypeError("argument 'base_dict' is not a ditionary type")
if not isinstance(upd_dict, dict):
raise TypeError("argument 'upd_dict' is not a ditionary type")
if names is None:
names = []
for upd_key, upd_val in upd_dict.items():
# For each upd_key & upd_val pair in upd_dict:
if isinstance(upd_val, dict):
if upd_key in base_dict:
# if base_dict[upd_key] exists, recursively deep-update it
if not isinstance(base_dict[upd_key], dict):
raise TypeError(f"type mismatch {'.'.join(names)}: is not a dict type in base_dict")
dict_deepupdate(
base_dict[upd_key],
upd_val,
names
+ [
upd_key,
],
)
else:
# if base_dict[upd_key] not exist, set base_dict[upd_key] from deepcopy of upd_val
base_dict[upd_key] = copy.deepcopy(upd_val)
elif isinstance(upd_val, list):
if upd_key in base_dict:
# if base_dict[upd_key] exists, base_dict[up_key] is extended by
# the list from upd_val
if not isinstance(base_dict[upd_key], list):
raise TypeError(f"type mismatch {'.'.join(names)}: is not a list type in base_dict")
base_dict[upd_key].extend(upd_val)
else:
# if base_dict[upd_key] doesn't exists, set base_dict[key] from a deepcopy of the
# list in upd_val.
base_dict[upd_key] = copy.deepcopy(upd_val)
elif isinstance(upd_val, set):
if upd_key in base_dict:
# if base_dict[upd_key] exists, base_dict[up_key] is updated by the set in upd_val
if not isinstance(base_dict[upd_key], set):
raise TypeError(f"type mismatch {'.'.join(names)}: is not a set type in base_dict")
base_dict[upd_key].update(upd_val.copy())
else:
# if base_dict[upd_key] doesn't exists, set base_dict[upd_key] from a copy of the
# set in upd_val
base_dict[upd_key] = upd_val.copy()
else:
# for any other type of upd_val replace or add base_dict[upd_key] by a copy
# of upd_val
base_dict[upd_key] = copy.copy(upd_val)

View file

@ -94,6 +94,7 @@ from searx.utils import (
from searx.version import VERSION_STRING, GIT_URL, GIT_BRANCH
from searx.query import RawTextQuery
from searx.plugins import Plugin, plugins, initialize as plugin_initialize
from searx.botdetection import link_token
from searx.plugins.oa_doi_rewrite import get_doi_resolver
from searx.preferences import (
Preferences,
@ -416,6 +417,7 @@ def render(template_name: str, **kwargs):
kwargs['endpoint'] = 'results' if 'q' in kwargs else request.endpoint
kwargs['cookies'] = request.cookies
kwargs['errors'] = request.errors
kwargs['link_token'] = link_token.get_token()
# values from the preferences
kwargs['preferences'] = request.preferences
@ -642,6 +644,12 @@ def health():
return Response('OK', mimetype='text/plain')
@app.route('/client<token>.css', methods=['GET', 'POST'])
def client_token(token=None):
link_token.ping(request, token)
return Response('', mimetype='text/css')
@app.route('/search', methods=['GET', 'POST'])
def search():
"""Search query in q and return results.

View file

@ -50,9 +50,13 @@ class SelfIPTest(SearxTestCase):
self.assertTrue(len(store.plugins) == 1)
# IP test
request = Mock(remote_addr='127.0.0.1')
request.headers.getlist.return_value = []
search = get_search_mock(query='ip', pageno=1)
request = Mock()
request.remote_addr = '127.0.0.1'
request.headers = {'X-Forwarded-For': '1.2.3.4, 127.0.0.1', 'X-Real-IP': '127.0.0.1'}
search = get_search_mock(
query='ip',
pageno=1,
)
store.call(store.plugins, 'post_search', request, search)
self.assertTrue('127.0.0.1' in search.result_container.answers["ip"]["answer"])
@ -62,7 +66,6 @@ class SelfIPTest(SearxTestCase):
# User agent test
request = Mock(user_agent='Mock')
request.headers.getlist.return_value = []
search = get_search_mock(query='user-agent', pageno=1)
store.call(store.plugins, 'post_search', request, search)
@ -98,7 +101,6 @@ class HashPluginTest(SearxTestCase):
self.assertTrue(len(store.plugins) == 1)
request = Mock(remote_addr='127.0.0.1')
request.headers.getlist.return_value = []
# MD5
search = get_search_mock(query='md5 test', pageno=1)