The GNU Name System
The GNU Name System (GNS) is secure and decentralized naming system. It allows its users to register names as top-level domains (TLDs) and resolve other namespaces within their TLDs.
GNS is designed to provide:
Secure name resolution
Compatibility with DNS
For the initial configuration and population of your GNS installation, please follow the GNS setup instructions. The remainder of this chapter will provide some background on GNS and then describe how to use GNS in more detail.
Unlike DNS, GNS does not rely on central root zones or authorities. Instead any user administers their own root and can can create arbitrary name value mappings. Furthermore users can delegate resolution to other users’ zones just like DNS NS records do. Zones are uniquely identified via public keys and resource records are signed using the corresponding public key. Delegation to another user’s zone is done using special PKEY records and petnames. A petname is a name that can be freely chosen by the user. This results in non-unique name-value mappings as www.bob to one user might be www.friend for someone else.
In the default configuration, there are two zones defined and shipped with GNUnet:
The first is “gnunet.org”, which points to the authoritate zone of the GNUnet project. It can be used to resolve, for example, “www.gnunet.org”.
“.pin” is another default zone which points to a special zone also managed by gnunet.org. Users may register submodomains on a first-come first-served-basis at https://fcfs.gnunet.org.
Use gnunet-config -s gns to view the GNS configuration, including all configured zones that are operated by other users. The respective configuration entry names start with a “.”, e.g. “.pin”.
You can configure any number of top-level domains, and point them to the respective zones of your friends! For this, simply obtain the respective public key (you will learn how below) and extend the configuration:
$ gnunet-config -s gns -o .myfriend -V PUBLIC_KEY
Zones and Egos
In GNUnet, identity management is about managing egos. Egos can correspond to pseudonyms or real-world identities. If you value your privacy, you are encouraged to use separate egos for separate activities.
Technically, an ego is first of all a public-private key pair, and thus egos also always correspond to a GNS zone. Egos are managed by the IDENTITY service. Note that this service has nothing to do with the peer identity. The IDENTITY service essentially stores the private keys under human-readable names, and keeps a mapping of which private key should be used for particular important system functions.
You probably should create at least one zone of your own. You can create any number of zones using the gnunet-identity tool using:
$ gnunet-identity --create="myzone"
Henceforth, on your system you control the TLD “myzone”.
All of your zones can be listed (displayed) using the gnunet-identity command line tool as well:
$ gnunet-identity --display
Now you can add (or edit, or remove) records in your GNS zone using the gnunet-namestore-gtk GUI or using the gnunet-namestore command-line tool. In either case, your records will be stored in an SQL database under control of the gnunet-service-namestore. Note that if multiple users use one peer, the namestore database will include the combined records of all users. However, users will not be able to see each other’s records if they are marked as private.
To provide a short example for editing your own zone, suppose you have your own web server with the IP 22.214.171.124. Then you can put an A record (A records in DNS are for IPv4 IP addresses) into your local zone “myzone” using the command:
$ gnunet-namestore -z myzone -a -n www -t A -V 126.96.36.199 -e 1d
Similar commands will work for other types of DNS and GNS records, the syntax largely depending on the type of the record. Naturally, most users may find editing the zones using the gnunet-namestore-gtk GUI to be easier.
Each zone in GNS has a public-private key. Usually, gnunet-namestore and gnunet-setup will access your private key as necessary, so you do not have to worry about those. What is important is your public key (or rather, the hash of your public key), as you will likely want to give it to others so that they can securely link to you.
A central operation in GNS is the ability to securely delegate to other zones. Basically, by adding a delegation you make all of the names from the other zone available to yourself. This section describes how to create delegations.
Suppose you have a friend who you call ’bob’ who also uses GNS. You can then delegate resolution of names to Bob’s zone by adding a PKEY record to their local zone:
$ gnunet-namestore -a -n bob --type PKEY -V XXXX -e 1d -z myzone
Note that “XXXX” in the command above must be replaced with the hash of Bob’s public key (the output your friend obtained using the gnunet-identity command from the previous section and told you, for example by giving you a business card containing this information as a QR code).
Assuming Bob has an “A” record for their website under the name of “www” in his zone, you can then access Bob’s website under “www.bob.myzone” — as well as any (public) GNS record that Bob has in their zone by replacing www with the respective name of the record in Bob’s zone.
Furthermore, if Bob has themselves a (public) delegation to Carol’s zone under “carol”, you can access Carol’s records under “NAME.carol.bob.myzone” (where “NAME” is the name of Carol’s record you want to access).
Resolving GNS records
Next, you should try resolving your own GNS records. The method we found to be the most uncomplicated is to do this by explicitly resolving using gnunet-gns. For this exercise, we will assume that you used the string “gnu” for the pseudonym (or label) of your GNS zone. If you used something else, replace “.gnu” with your real pseudonym in the examples below.
In the shell, type:
$ gnunet-gns -u www.gnunet.org www.gnunet.org: Got `A' record: ...
That shows that resolution works, once GNS is integrated with the application.
Integration with Browsers (DNS2GNS service)
Most OSes allow you to either modify your
resolvectl. We are going to configure the dns2gns service
in order to translate DNS name queries by applications to GNS name
queries where applicable and else fall back to DNS.
Optionally, you may want to configure your dns2gns service to run on a
non-priviledged port like 5353. But, in case you are going to edit
/etc/resolv.conf directly, the dns2gns service MUST run on port 53
as you cannot specify the port number. A
should be a DNS server you trust such as your local router:
$ gnunet-config -s dns2gns -o OPTIONS -V "-d $FALLBACK_DNS -p 5252" $ gnunet-arm -i dns2gns # Make sure the service is started
If you edit your resolv.conf directly, it should contain and entry like this:
In any case, it is very likely that the method of modification of your resolver is OS specific. Recently, the combination of NetworkManager and systemd-resolved is becoming increasingly popular.
If you use resolvectl and systemd-resolved you can temporarily set the nameserver like this:
$ resolvectl $INTERFACE 127.0.0.1:5353
$INTERFACE is your network interface such as
In order to automatically set the DNS2GNS server if it is running already you can use NetworkManager-dispatcher. First, enable it:
$ sudo systemctl enable NetworkManager-dispatcher.service $ sudo systemctl start NetworkManager-dispatcher.service
Then, create a script
#!/bin/sh interface=$1 status=$2 if [ "$interface" = "eth0" ]; then case $status in up) if nc -u -z 127.0.0.1 5353; then resolvectl dns $interface 127.0.0.1:5353 fi ;; down) ;; esac fi
Make sure the script is owned by root and executable:
$ sudo root:root /etc/NetworkManager/dispatch.d/10-dns2gns.sh $ sudo +x /etc/NetworkManager/dispatch.d/10-dns2gns.sh
You can test accessing this website using your browser or curl:
$ curl www.gnunet.org
Note that “gnunet.org” is a domain that also exists in DNS and for which the GNUnet project webservers can provide trusted TLS certificates. When using non-DNS names with GNS or aliases, this may result in issues when accessing HTTPS websites with browsers. In order learn how to provide relief for this issue, read on.
Integration with Browsers (SOCKS proxy)
While we recommend integrating GNS using the DNS2GNS service or the NSSwitch plugin, you can also integrate GNS directly with your browser via the gnunet-gns-proxy. This method can have the advantage that the proxy can validate TLS/X.509 records and thus strengthen web security; however, the proxy is still a bit brittle, so expect subtle failures. We have had reasonable success with Chromium, and various frustrations with Firefox in this area recently.
The first step is to start the proxy. As the proxy is (usually) not started by default, this is done as a unprivileged user using gnunet-arm -i gns-proxy. Use gnunet-arm -I as a unprivileged user to check that the proxy was actually started. (The most common error for why the proxy may fail to start is that you did not run gnunet-gns-proxy-setup-ca during installation.) The proxy is a SOCKS5 proxy running (by default) on port 7777. Thus, you need to now configure your browser to use this proxy. With Chromium, you can do this by starting the browser as a unprivileged user using chromium –proxy-server=“socks5://localhost:7777” For Firefox (or Icecat), select “Edit-Preferences” in the menu, and then select the “Advanced” tab in the dialog and then “Network”:
Here, select “Settings…” to open the proxy settings dialog. Select “Manual proxy configuration” and enter localhost with port 7777 under SOCKS Host. Furthermore, set the checkbox “Proxy DNS when using SOCKS v5” at the bottom of the dialog. Finally, push “OK”.
You must also go to about:config and change the browser.fixup.alternate.enabled option to false, otherwise the browser will autoblunder an address like www.gnu to www.gnu.com. If you want to resolve @ in your own TLDs, you must additionally set browser.fixup.dns_first_use_for_single_words to true.
After configuring your browser, you might want to first confirm that it continues to work as before. (The proxy is still experimental and if you experience “odd” failures with some webpages, you might want to disable it again temporarily.) Next, test if things work by typing “http://test.gnu/” into the URL bar of your browser. This currently fails with (my version of) Firefox as Firefox is super-smart and tries to resolve “http://www.test.gnu/” instead of “test.gnu”. Chromium can be convinced to comply if you explicitly include the “http://” prefix — otherwise a Google search might be attempted, which is not what you want. If successful, you should see a simple website.
Note that while you can use GNS to access ordinary websites, this is more an experimental feature and not really our primary goal at this time. Still, it is a possible use-case and we welcome help with testing and development.
Creating a Business Card
Before we can really use GNS, you should create a business card. Note that this requires having LaTeX installed on your system. If you are using a Debian GNU/Linux based operating system, the following command should install the required components. Keep in mind that this requires 3GB of downloaded data and possibly even more when unpacked. On a GNU Guix based system texlive 2017 has returns a DAG size of 5032.4 MiB. The packages which are confirmed to be required are:
We welcome any help in identifying the required components of the TexLive Distribution. This way we could just state the required components without pulling in the full distribution of TexLive.
apt-get install texlive-full
Start creating a business card by clicking the “Copy” button in gnunet-namestore-gtk. Next, you should start the gnunet-bcd program (in the terminal, on the command-line). You do not need to pass any options, and please be not surprised if there is no output:
$ gnunet-bcd # does not return
Then, start a browser and point it to http://localhost:8888/ where gnunet-bcd is running a Web server!
First, you might want to fill in the “GNS Public Key” field by right-clicking and selecting “Paste”, filling in the public key from the copy you made in gnunet-namestore-gtk. Then, fill in all of the other fields, including your GNS NICKname. Adding a GPG fingerprint is optional. Once finished, click “Submit Query”. If your LaTeX installation is incomplete, the result will be disappointing. Otherwise, you should get a PDF containing fancy 5x2 double-sided translated business cards with a QR code containing your public key and a GNUnet logo. We’ll explain how to use those a bit later. You can now go back to the shell running gnunet-bcd and press CTRL-C to shut down the Web server.
Backup of Identities and Egos
One should always backup their files, especially in these SSD days (our team has suffered 3 SSD crashes over a span of 2 weeks). Backing up peer identity and zones is achieved by copying the following files:
The peer identity file can be found in
The private keys of your egos are stored in the directory
~/.local/share/gnunet/identity/egos/. They are stored in files whose
filenames correspond to the zones’ ego names. These are probably the
most important files you want to backup from a GNUnet installation.
Note: All these files contain cryptographic keys and they are stored without any encryption. So it is advisable to backup encrypted copies of them.
Now, in the situation of an attacker gaining access to the private key of one of your egos, the attacker can create records in the respective GNS zone and publish them as if you published them. Anyone resolving your domain will get these new records and when they verify they seem authentic because the attacker has signed them with your key.
To address this potential security issue, you can pre-compute a revocation certificate corresponding to your ego. This certificate, when published on the P2P network, flags your private key as invalid, and all further resolutions or other checks involving the key will fail.
A revocation certificate is thus a useful tool when things go out of control, but at the same time it should be stored securely. Generation of the revocation certificate for a zone can be done through gnunet-revocation. For example, the following command (as unprivileged user) generates a revocation file revocation.dat for the zone zone1: gnunet-revocation -f revocation.dat -R zone1
The above command only pre-computes a revocation certificate. It does not revoke the given zone. Pre-computing a revocation certificate involves computing a proof-of-work and hence may take up to 4 to 5 days on a modern processor. Note that you can abort and resume the calculation at any time. Also, even if you did not finish the calculation, the resulting file will contain the signature, which is sufficient to complete the revocation process even without access to the private key. So instead of waiting for a few days, you can just abort with CTRL-C, backup the revocation certificate and run the calculation only if your key actually was compromised. This has the disadvantage of revocation taking longer after the incident, but the advantage of saving a significant amount of energy. So unless you believe that a key compromise will need a rapid response, we urge you to wait with generating the revocation certificate. Also, the calculation is deliberately expensive, to deter people from doing this just for fun (as the actual revocation operation is expensive for the network, not for the peer performing the revocation).
To avoid TL;DR ones from accidentally revocating their zones, we are not giving away the command, but it is uncomplicated: the actual revocation is performed by using the -p option of gnunet-revocation.
This may seem not like much of an application yet, but you have just been one of the first to perform a decentralized secure name lookup (where nobody could have altered the value supplied by your friend) in a privacy-preserving manner (your query on the network and the corresponding response were always encrypted). So what can you really do with this? Well, to start with, you can publish your GnuPG fingerprint in GNS as a “CERT” record and replace the public web-of-trust with its complicated trust model with explicit names and privacy-preserving resolution. Also, you should read the next chapter of the tutorial and learn how to use GNS to have a private conversation with your friend. Finally, help us with the next GNUnet release for even more applications using this new public key infrastructure.
GNS supports the majority of the DNS records as defined in RFC 1035. Additionally, GNS defines some new record types the are unique to the GNS system. For example, GNS-specific resource records are used to give petnames for zone delegation, revoke zone keys and provide some compatibility features.
For some DNS records, GNS does extended processing to increase their usefulness in GNS. In particular, GNS introduces special names referred to as “zone relative names”. Zone relative names are allowed in some resource record types (for example, in NS and CNAME records) and can also be used in links on webpages. Zone relative names end in “.+” which indicates that the name needs to be resolved relative to the current authoritative zone. The extended processing of those names will expand the “.+” with the correct delegation chain to the authoritative zone (replacing “.+” with the name of the location where the name was encountered) and hence generate a valid GNS name.
The GNS currently supports the record types as defined in GANA. In addition, GNS supports DNS record types, such as A, AAAA or TXT.
For a complete description of the records, please refer to the specification at LSD0001.
In the following, we discuss GNS records with specific behaviour or special handling in GNUnet.
GNS allows easy access to services provided by the GNUnet Virtual Public Network. When the GNS resolver encounters a VPN record it will contact the VPN service to try and allocate an IPv4/v6 address (if the queries record type is an IP address) that can be used to contact the service.
I want to provide access to the VPN service “web.gnu.” on port 80 on peer ABC012: Name: www; RRType: VPN; Value: 80 ABC012 web.gnu.
The peer ABC012 is configured to provide an exit point for the service “web.gnu.” on port 80 to it’s server running locally on port 8080 by having the following lines in the gnunet.conf configuration file:
[web.gnunet.] TCP_REDIRECTS = 80:localhost4:8080
Synchronizing with legacy DNS
If you want to support GNS but the master database for a zone is only available and maintained in DNS, GNUnet includes the gnunet-zoneimport tool to monitor a DNS zone and automatically import records into GNS. Today, the tool does not yet support DNS AF(X)R, as we initially used it on the “.fr” zone which does not allow us to perform a DNS zone transfer. Instead, gnunet-zoneimport reads a list of DNS domain names from stdin, issues DNS queries for each, converts the obtained records (if possible) and stores the result in the namestore.
The zonemaster service then takes the records from the namestore, publishes them into the DHT which makes the result available to the GNS resolver. In the GNS configuration, non-local zones can be configured to be intercepted by specifying “.tld = PUBLICKEY” in the configuration file in the “[gns]” section.
Note that the namestore by default also populates the namecache. This pre-population is cryptographically expensive. Thus, on systems that only serve to import a large (millions of records) DNS zone and that do not have a local gns service in use, it is thus advisable to disable the namecache by setting the option “DISABLE” to “YES” in section “[namecache]”.
Migrating an existing DNS zone into GNS
Ascension is a tool to migrate existing DNS zones into GNS.
Compared to the gnunet-zoneimport tool it strictly uses AXFR or IXFR depending on whether or not there exists a SOA record for the zone. If that is the case it will take the serial as a reference point and request the zone. The server will either answer the IXFR request with a correct incremental zone or with the entire zone, which depends on the server configuration.
Before you can migrate any zone though, you need to start a local GNUnet peer. To migrate the Syrian top level domain - one of the few top level domains that support zone transfers - into GNS use the following command:
$ ascension sy. -n ns1.tld.sy. -p
The -p flag will tell GNS to put these records on the DHT so that other users may resolve these records by using the public key of the zone.
Once the zone is migrated, Ascension will output a message telling you, that it will refresh the zone after the time has elapsed. You can resolve the names in the zone directly using GNS or if you want to use it with your browser, check out the GNS manual section. Configuring the GNU Name System. To resolve the records from another system you need the respective zones PKEY. To get the zones public key, you can run the following command:
$ gnunet-identity -dqe sy
Where “sy” is the name of the zone you want to migrate.
You can share the PKEY of the zone with your friends. They can then resolve records in the zone by doing a lookup replacing the zone label with your PKEY:
$ gnunet-gns -t SOA -u "$PKEY"
The program will continue to run as a daemon and update once the refresh time specified in the zones SOA record has elapsed.
DNSCurve style records are supported in the latest release and they are added as a PKEY record to be referred to the respective GNS public key. Key distribution is still a problem but provided someone else has a public key under a given label it can be looked up.
There is an unofficial Debian package called python3-ascension that adds a system user ascension and runs a GNUnet peer in the background.
Ascension-bind is also an unofficial Debian package that on installation checks for running DNS zones and whether or not they are transferable using DNS zone transfer (AXFR). It asks the administrator which zones to migrate into GNS and installs a systemd unit file to keep the zone up to date. If you want to migrate different zones you might want to check the unit file from the package as a guide.