commit | a7f280aee3fb6a3164db4cc8ec32cf4303b96b46 | [log] [tgz] |
---|---|---|
author | Daniel Gruno <humbedooh@apache.org> | Wed Oct 19 16:17:10 2022 -0500 |
committer | Daniel Gruno <humbedooh@apache.org> | Wed Oct 19 16:17:10 2022 -0500 |
tree | b775509f880a893934e3c5011651e5c06a7c1d3b | |
parent | 2d77234b540c4adab34da2e8c461a7a8f5d734dc [diff] |
add a note on header dupes
Aardvark acts as a middleman between frontend web servers and (typically) ticket submission services such as JIRA or BugZilla, and intercepts all data sent. POST Data is scanned for known offending words that are common in spam, and if found to be spam, the request is blocked. Aardvark keeps an internal list of offending IPs, and will block any subsequent POST requests from those IPs (until restarted).
Aardvark is written in Python3 and uses aiohttp for its server/client capabilities.
port
: Which port to listen on for scans. For security reasons, Aardvark will bind to localhost. Default is 1729proxy_url
: The backend service to proxy to if request is saneipheader
: The header to look for the client's IP in. Typically X-Forwarded-For.debug
: If set to true
, will spit out some extra lines for reach request handled. Can get very spammy.naive_spam_threshold
: This is the spam score threshold for the naïve scanner, spamfilter.py
. It uses a pre-generated English corpus for detecting spam.spamurls
: Specific honey-pot URLs that trigger a block regardless of the actionignoreurls
: Specific URLs that are exempt from spam detectionpostmatches
: A list of keywords and/or regexes that, if matched, will block the requestmultimatch
: A combination blocker. If a required
keyword or regex is matched, the request will be blocked only if one or more auxiliary
keywords/regexes are also matchedpersistence
: Enables persistent storage of offending IPs in blocklist.txt
. Enabling this also enables you to use unblock.py (to be enhanced further at a later point).savedata
: A path which, if set, is where debug data from offending requests will be saved. This is typically the full first request an IP makes.suppress_repeats
: Suppresses repeat syslog entries for known offenders. debug: true
will override this.Aardvark contains a very naïve spam scanner in spamfilter.py
that uses a very simplified Bayes-esque formula for determining whether something is spam. It is enabled for form data only, and can be disabled entirely by setting enable_naive_scan
to false
. It has a built-in corpus with ham and spam in English, and works...sometimes :)
It is very much a work in progress, but should be safe to have enabled.
To enable as a pipservice, add the following minimal hiera yaml to your node config:
pipservice: aardvark-proxy: tag: main
Follow these steps to run manually (assuming you have pipenv installed):
git clone https://github.com/apache/infrastructure-aardvark-proxy.git aardvark-proxy
cd aardvark-proxy
pipenv install -r requirements.txt
pipenv run python3 aardvark.py
As Aardvark is a proxy middleman for specific purposes, you will preferably need a web server in front. The example below relays all POST requests for /foo/bar through Aardvark, while letting all GETs etc go directly to the backend service.
Assuming Aardvark is listening on port 1729
and the real backend service is on port 8080
:
# Send all POST requests through Aardvark RewriteEngine On RewriteCond %{REQUEST_METHOD} POST RewriteRule ^/(.*)$ http://localhost:1729/$1 [P] # Rest goes to backend directly ProxyPass / http://localhost:8080/foo/bar/
IPs can be unblocked in a couple of ways:
python3 unblock.py ip.goes.here
curl 'http://localhost:1729/aardvark-unblock?ip.goes.here' -H 'X-Aardvark-Key: uuid-from-blocklist.txt'