AdHole is a simple transparent advertisement and tracking blocker intended for personal use.
It works by providing two elements: first a DNS server that, based on a list of domains, either relies the query to a real server, or returns its own IP address; and a micro HTTP server that serves an empty GIF image for any request. As such its functionality is akin to AdBlock , but less pretty (you'll probably see empty space where ads used to be) and completely transparent (i.e. will work for any browser, on any OS on any machine, locally or over any network). Note that you can just as well block any domains.
I use AdHole in 'production' on my RaspberryPi-powered WiFi router, where it provides ad and tracking blocking to all my wireless toys.
My programming goal was to make it as simple and minimal as possible. Therefore
instead of a full DNS stack (and especially parsing) I work directly on
[]byte
slices and try to reuse the data already on hand. It was also a nice
little adventure in dealing with binary protocols the low-level way.
- 2016-11-11 Added locking around the queries map. Can't really test it as the RPi is obviously single-core.
On a Unix-like system you should be able to get everything built using just:
$ make
cd adhole; \
gofmt -w *.go; \
go build .
cd genlist; \
gofmt -w *.go; \
go build .
Otherwise just run go build .
in either or both of adhole/
and genlist/
.
$ ./adhole
Usage: ./adhole [options] key upstream proxy list.txt
key - password used for /debug actions protection
upstream - real upstream DNS address, e.g. 8.8.8.8
proxy - servers' bind address, e.g. 127.0.0.1
list.txt - text file with domains to block
-dport=53: DNS server port
-hport=80: HTTP server port
-t=5s: upstream query timeout
-v=false: be verbose
Note that you will need root privileges to run it on the default ports.
List format is simply: one domain name per line. All subdomains of a given
domain will be blocked, so there is no need to use *
. Domains should also not
end with a dot. The parser should also be indifferent to line endings. Example
list file:
101com.com
101order.com
103bees.com
123found.com
123pagerank.com
To get a decent list of domains to block I recommend going here and generating a 'plain non-HTML list -- as a plain list of hostnames (no HTML)' with 'no links back to this page' and 'view list as plain text:' ticked (that was so verbose...). Or alternatively this should work as a direct link to the current list...
Or you can use the experimental genlist
utility. Currently I've implemented
fetching for the above list and for
EasyList.
If you have a better / your favourite source of domain blacklist
please look at sources.go
and implement a fetcher. As a side note, sources
that are meant to be used by AdBlock (or equivalent) are not the best
candidates here, due to the fact that AdBlock is much more subtle, i.e. with
current extraction method a list generated with EasyList input will block
google.com, bbc.co.uk and imdb.com (and probably many more sites that you
care about). Also the whole thing is brittle by nature, hence the permanent
'experimental' status.
$ ./genlist
Usage: ./genlist [options] list|all|source source source...
list - show available sources
all - combine all sources
source - use only specified source(s)
$ ./genlist list
There are 2 sources available:
01 - pgl - http://pgl.yoyo.org/adservers
02 - easylist - https://easylist.adblockplus.org
$ ./genlist all > /tmp/blacklist.txt
Processing list for pgl
Processing list for easylist
Got 9829 domains total
Example systemd service file:
[Unit]
Description=DNS blackhole
After=network.target
[Service]
ExecStart=/home/user/adhole SecretKey 192.168.0.3 192.168.0.21 /home/user/blacklist.txt
[Install]
WantedBy=multi-user.target
Thanks to the great expvar package you can
monitor some statistics by visiting http://proxy.addr/debug/vars
. The
following items are relevant:
stateIsRunning
- if false all queries are relied to upstreamstatsQuestions
- number of received queriesstatsRelayed
- number of queries relayed to the real serverstatsBlocked
- number of queries blockedstatsTimedout
- number of relayed queries that timed outstatsServed
- number of HTTP requests servedstatsErrors
- number of errors encounteredstatsRules
- number of items read from the blacklist
You can also do the following actions via HTTP:
/debug/reload
- will reload the list.txt file/debug/toggle
- toggle blocking on and off
You'll need to append &key=YOURKEY
to the above. Unauthorized hits will
be logged. Note that you may set the key to ""
(i.e. an empty key) and
therefore disable the authentication.
Tested on:
- Linux - amd64, armv6l
- Windows XP - i686
Feel free to either use GitHub's pull request or send me patches directly.
I also welcome any feedback regarding stability and what OS/platform you run AdHole on (just out of curiosity).
- Edge cases (multiple questions per query, anybody?)
- Even less data shuffling
- IPv4 only
Copyright (c) 2014 - 2019 Piotr S. Staszewski
Absolutely no warranty. See LICENSE.txt for details.