A protocol for securing communications between a client and a DNS resolver.
dnscrypt-proxy
verifies that responses you get from a DNS provider have been
actually sent by that provider, and haven't been tampered with.
This is not a VPN. It doesn't mask your IP address, and if you are using it with a public DNS service, be aware that it will (and has to) decrypt your queries.
If you are using it for privacy, it might do the opposite of what you are trying to achieve. It you are using it to prevent VPN "leaks", this isn't the right tool either: the proper way to prevent VPN "leaks" is to avoid sending data to yet another third party: use a VPN service that operates its own DNS resolvers.
dnscrypt-proxy
provides local service which can be used directly as
your local resolver or as a DNS forwarder, authenticating requests
using the DNSCrypt protocol and passing them to an upstream server.
The DNSCrypt protocol uses high-speed high-security elliptic-curve cryptography and is very similar to DNSCurve, but focuses on securing communications between a client and its first-level resolver.
While not providing end-to-end security, it protects the local network, which is often the weakest point of the chain, against man-in-the-middle attacks.
dnscrypt-proxy can be downloaded here: dnscrypt-proxy download
Note: dnscrypt.org is now blocked by the Great Firewall of China. But the source code can also be downloaded on Github, in the "releases" section.
After having downloaded a file, compute its SHA256 digest. For example:
$ openssl dgst -sha256 dnscrypt-proxy-1.6.0.tar.bz2
Verify this digest against the expected one, that can be retrieved using a simple DNS query:
$ drill -D TXT dnscrypt-proxy-1.6.0.tar.bz2.download.dnscrypt.org
or
$ dig +dnssec TXT dnscrypt-proxy-1.6.0.tar.bz2.download.dnscrypt.org
If the content of the TXT record doesn't match the SHA256 digest you computed, please file a bug report on Github as soon as possible and don't go any further.
The daemon is known to work on recent versions of OSX, OpenBSD, Bitrig, NetBSD, Dragonfly BSD, FreeBSD, Linux, iOS (requires a jailbroken device), Android (requires a rooted device), and Windows (requires MingW).
Install libsodium.
On Linux, don't forget to run ldconfig
if you installed it from
source.
A "minimal" build of libsodium (--enable-minimal
) works equally well
as a full build with this proxy.
On Fedora, RHEL and CentOS, you may need to add /usr/local/lib
to
the paths the dynamic linker is going to look at. Before issuing
ldconfig
, type:
# echo /usr/local/lib > /etc/ld.so.conf.d/usr_local_lib.conf
Now, download the latest dnscrypt-proxy
version and extract it:
$ bunzip2 -cd dnscrypt-proxy-*.tar.bz2 | tar xvf -
$ cd dnscrypt-proxy-*
Compile and install it using the standard procedure:
$ ./configure && make -j2
# make install
Replace -j2
with whatever number of CPU cores you want to use for the
compilation process.
The proxy will be installed as /usr/local/sbin/dnscrypt-proxy
by default.
Command-line switches are documented in the dnscrypt-proxy(8)
man page.
Note: gcc 3.4.6 (and probably other similar versions) is known to produce broken code on Mips targets with the -Os optimization level. Use a different level (-O and -O2 are fine) or upgrade the compiler. Thanks to Adrian Kotelba for reporting this.
If you need a simple graphical user interface in order to start/stop the proxy and change your DNS settings, check out the following project:
-
DNSCrypt WinClient: Easily enable/disable DNSCrypt on multiple adapters. Supports different ports and protocols, IPv6, parental controls and the proxy can act as a gateway service. Windows only, written in .NET.
-
DNSCrypt Windows Service Manager: Assists in setting up DNSCrypt as a service, configure it and change network adapter DNS settings to use DNSCrypt. It includes the option to use TCP/UDP protocol, IPV4/IPV6 connectivity, choice of network adapter to configure, as well as configurations for currently available DNSCrypt providers.
-
DNSCrypt OSXClient: Mac OSX application to control the DNSCrypt Proxy.
-
DNSCrypt Tools for Linux: A set of tools for
dnscrypt-proxy
. Features a start and stop button as well as options to enable or disable from startup. Developed for Porteus Linux.
To get started, you can use any of the public DNS resolvers supporting DNSCrypt.
This file is constantly updated, and its minisign signature can be verified with the following command:
minisign -VP RWQf6LRCGA9i53mlYecO4IzT51TGPpvWucNSCh1CBM0QTaLn73Y7GFO3 -m dnscrypt-resolvers.csv
If you want to add DNSCrypt support to your own public or private resolver, check out DNSCrypt-Wrapper, a server-side dnscrypt proxy that works with any name resolver.
A DNSCrypt server Docker image is also available to deploy a non-logging, DNSSEC and DNSCrypt-capable resolver without having to manually compile or configure anything.
Having a dedicated system user, with no privileges and with an empty home directory, is highly recommended. For extra security, DNSCrypt will chroot() to this user's home directory and drop root privileges for this user's uid as soon as possible.
The easiest way to start the daemon is:
# dnscrypt-proxy --daemonize --resolver-name=<resolver name>
Replace <resolver name>
with the name of the resolver you want to
use (the first column in the list of public resolvers).
The proxy will accept incoming requests on 127.0.0.1, tag them with an authentication code, forward them to the resolver, and validate each answer before passing it to the client.
Given such a setup, in order to actually start using DNSCrypt, you
need to update your /etc/resolv.conf
file and replace your current
set of resolvers with:
nameserver 127.0.0.1
Other common command-line switches include:
--daemonize
in order to run the server as a background process.--local-address=<ip>[:port]
in order to locally bind a different IP address than 127.0.0.1--logfile=<file>
in order to write log data to a dedicated file. By default, logs are sent to stdout if the server is running in foreground, and to syslog if it is running in background.--loglevel=<level>
if you need less verbosity in log files.--max-active-requests=<count>
to set the maximum number of active requests. The default value is 250.--pidfile=<file>
in order to store the PID number to a file.--user=<user name>
in order to chroot()/drop privileges.--resolvers-list=<file>
: to specity the path to the CSV file containing the list of available resolvers, and the parameters to use them.--test
in order to check that the server-side proxy is properly configured and that a valid certificate can be used. This is useful for monitoring your own dnscrypt proxy. See the man page for more information.
The
--resolver-address=<ip>[:port]
,
--provider-name=<certificate provider FQDN>
and --provider-key=<provider public key>
switches can be specified in
order to use a DNSCrypt-enabled recursive DNS service not listed in
the configuration file.
On a system using systemd, and when compiled with --with-systemd
,
the proxy can take advantage of systemd's socket activation instead of
creating the sockets itself. The proxy will also notify systemd on successful
startup.
Two sockets need to be configured: a UDP socket (ListenStream
) and a
TCP socket (ListenDatagram
) sharing the same port.
The source distribution includes the dnscrypt-proxy.socket
and
dnscrypt-proxy.service
files that can be used as a starting point.
The proxy can be installed as a Windows service.
See README-WINDOWS.markdown for more information on DNSCrypt on Windows.
The DNSCrypt proxy is not a DNS cache. This means that incoming queries will not be cached and every single query will require a round-trip to the upstream resolver.
For optimal performance, the recommended way of running DNSCrypt is to
run it as a forwarder for a local DNS cache, such as unbound
or
powerdns-recursor
.
Both can safely run on the same machine as long as they are listening to different IP addresses (preferred) or different ports.
If your DNS cache is unbound
, all you need is to edit the
unbound.conf
file and add the following lines at the end of the server
section:
do-not-query-localhost: no
forward-zone:
name: "."
forward-addr: 127.0.0.1@40
The first line is not required if you are using different IP addresses instead of different ports.
Then start dnscrypt-proxy
, telling it to use a specific port (40
, in
this example):
# dnscrypt-proxy --local-address=127.0.0.1:40 --daemonize
IPv6 is fully supported. IPv6 addresses with a port number should be specified as [ip]:port
# dnscrypt-proxy --local-address='[::1]:40' ...
Some routers and firewalls can block outgoing DNS queries or transparently redirect them to their own resolver. This especially happens on public Wifi hotspots, such as coffee shops.
As a workaround, the port number can be changed using
the --resolver-port=<port>
option.
By default, dnscrypt-proxy
sends outgoing queries to UDP port 443.
In addition, the DNSCrypt proxy can force outgoing queries to be sent over TCP. For example, TCP port 443, which is commonly used for communication over HTTPS, may not be filtered.
The --tcp-only
command-line switch forces this behavior. When
an incoming query is received, the daemon immediately replies with a
"response truncated" message, forcing the client to retry over TCP.
The daemon then authenticates the query and forwards it over TCP
to the resolver.
--tcp-only
is slower than UDP because multiple queries over a single
TCP connections aren't supported yet, and this workaround should
never be used except when bypassing a filter is actually required.
By default, dnscrypt-proxy generates non-deterministic client keys every time it starts, or for every query (when the ephemeral keys feature is turned on).
However, commercial DNS services may want to use DNSCrypt to authenticate the sender of a query using public-key cryptography, i.e. know what customer sent a query without altering the DNS query itself, and without using shared secrets.
Resolvers that should be accessible from any IP address, but that are supposed to be used only by specific users, can also take advantage of DNSCrypt to only respond to queries sent using a given list of public keys.
In order to do so, dnscrypt-proxy 1.6.0 introduced the --client-key
(or -K
) switch. This loads a secret client key from a file instead
of generating random keys:
# dnscrypt-proxy --client-key=/private/client-secret.key
This file has to remain private, and its content doesn't have to be known by the DNS service provider.
Versions 1 and 2 of the DNSCrypt protocol use Curve25519 keys, and the
format of this file for Curve25519 keys is a hexadecimal string, with
optional :
, [space]
and -
delimiters, decoding to 34 bytes:
01 01 || 32-byte Curve25519 secret key
DNS packets sent over UDP have been historically limited to 512 bytes, which is usually fine for queries, but sometimes a bit short for replies.
Most modern authoritative servers, resolvers and stub resolvers support the Extension Mechanism for DNS (EDNS) that, among other things, allows a client to specify how large a reply over UDP can be.
Unfortunately, this feature is disabled by default on a lot of
operating systems. It has to be explicitly enabled, for example by
adding options edns0
to the /etc/resolv.conf
file on most
Unix-like operating systems.
dnscrypt-proxy
can transparently rewrite outgoing packets before
authenticating them, in order to add the EDNS0 mechanism. By
default, a conservative payload size of 1252 bytes is advertised.
This size can be made larger by starting the proxy with the
--edns-payload-size=<bytes>
command-line switch. Values up to 4096
are usually safe, but some routers/firewall/NAT boxes block IP fragments.
If you can resolve test-tcp.dnscrypt.org
, increasing the maximum
payload size is probably fine. If you can't, or just to stay on the
safe side, do not tweak this; stick to the default value.
A value below or equal to 512 will disable this mechanism, unless a client sends a packet with an OPT section providing a payload size.
The DNSCrypt proxy ships with a simple tool named hostip
that
resolves a name to IPv4 or IPv6 addresses.
This tool can be useful for starting some services before
dnscrypt-proxy
.
Queries made by hostip
are not authenticated.
dnscrypt-proxy
can be extended with plugins. A plugin acts as a
filter that can locally inspect and modify queries and responses.
The plugin API is documented in the README-PLUGINS.markdown
file.
Any number of plugins can be combined (chained) by repeating the
--plugin
command-line switch.
The default distribution ships with some example plugins:
libdcplugin_example_ldns_aaaa_blocking
: Directly return an empty response to AAAA queries
Example usage:
# dnscrypt-proxy ... \
--plugin libdcplugin_example_ldns_aaaa_blocking.la
If IPv6 connectivity is not available on your network, this plugin avoids waiting for responses about IPv6 addresses from upstream resolvers. This can improve your web browsing experience.
libdcplugin_example_ldns_blocking
: Block specific domains and IP addresses.
This plugin returns a REFUSED response if the query name is in a list of blacklisted names, or if at least one of the returned IP addresses happens to be in a list of blacklisted IPs.
Recognized switches are:
--domains=<file>
--ips=<file>
A file should list one entry per line.
IPv4 and IPv6 addresses are supported.
For names, leading and trailing wildcards (*
) are also supported
(e.g. *xxx*
, *.example.com
, ads.*
)
# dnscrypt-proxy ... \
--plugin libdcplugin_example,--ips=/etc/blk-ips,--domains=/etc/blk-names
libdcplugin_example-logging
: Log client queries
This plugin logs the client queries to the standard output (default) or to a file.
# dnscrypt-proxy ... \
--plugin libdcplugin_example_logging,/var/log/dns.log
- Extra plugins
Additional plugins can be found on Github: Masquerade plugin, GeoIP plugin.