Update: Added some remarks about what DNS RPZ actually is, what my objective is and what the outcome will be.

A friend of mine and I tried to play w/ RPZ and knot yesterday and gravely failed. The fact that knot as well as RPZ had been new to us didn’t help. Discussing the failure later that night I remembered that I was already doing something similar at home for adblocking at the DNS level instead of every application on every client. In some way this is also DNS RPZ.

DNS RPZ is something that could be described as DNS firewalling and is described by wikipedia as follows:

Domain Name Service Response Policy Zones (DNS RPZ) is a method that allows a nameserver administrator to overlay custom information on top of the global DNS to provide alternate responses to queries. It is currently implemented in the ISC BIND nameserver (9.8 or later). Another generic name for the DNS RPZ functionality is “DNS firewall”.

My main objective is to block ad networks and malware sites (e.g. command and control) on the DNS level for all devices without them having to install adblockers or stuff. So if someone tries to access a blacklisted site, say domain.tld the client will get a NXDOMAIN as an answer instead of the real IP address. As I can’t keep up with the domains I also want to leverage some of the OSINT feeds available. Currently I have roughly 16k domains blacklisted.

So today I was incorporating some of my OSINT and PF work into my DNS setup at home. Just as a quick primer: I am using OpenBSD as the OS, unbound(8) as recursor and rcs(1) for local and simple version control. Nothing mentioned here is specific to OpenBSD besides the ftp(1) command that I use to fetch the feeds and the location of the files.

Setting up unbound

Setup is pretty simple, just add the following line to /var/unbound/etc/unbound.conf within the server: section of the file:

include: /var/unbound/zones/rpz

Then setup /var/unbound/zones/rpz w/ additional includes, e.g.

include: …

I came up with the idea to utilize Open Source Threat Feeds, or OSINT on my private setup and quickly cooked up the shell script below in a rough, first try. The funny thing is that I more or less instantantly got hits from the 5346 IP addresses in the table:

@0 block drop log quick from <pf_osint:5346> to any
  [ Evaluations: 502       Packets: 20        Bytes: 800         States: 0     ]
  [ Inserted: uid 0 pid 68515 State Creations: 0     ]

🧐. The script is currently designed to run every 3 hours and to remove all IP addresses which had been added to the table after a day. I am not too happy about some of the parts right now, first of all: I’d like to seperate download and parsing out to a non-privileged process and feed one large junk of data to pfctl(8).

This will give me the flexibility to replace the table with the new values so I can act fast not only on additions but also on deletions from the list. No need to keep out somebody longer than neccessary.

Not sure on how to attack the first part. A shell script would be easier and faster to create. Privsep could be handled by doas for the part needing higher privileges and allow me to handle everything in a single script.

But as a start I am going to keep the script as is and see how it’s going to work out for me. So without further ado, here’s the script:

printf "zeustracker.abuse.ch badips "
ftp  -VMo- https://zeustracker.abuse.ch/blocklist.php?download=badips | \
        egrep "^([0-9]{1,3}\.){3}[0-9]{1,3}$" | \
        pfctl -t pf_osint -T add -f -

printf "feodotracker.abuse.ch ipblocklist "
ftp  -VMo- https://feodotracker.abuse.ch/blocklist/?download=ipblocklist | \
        egrep "^([0-9]{1,3}\.){3}[0-9]{1,3}$" | \
        pfctl -t pf_osint -T add -f -

printf "ransomwaretracker.abuse.ch "
ftp -VMo- https://ransomwaretracker.abuse.ch/feeds/csv/ | \
        egrep -o '([1-9]{1,3}\.){3}[1-9]{1,3}[^/"]*' | \
        tr '|' …