A tool for securing communications between a client and a DNS resolver.
DNSCrypt is a slight variation on DNSCurve.
DNSCurve improves the confidentiality and integrity of DNS requests using high-speed high-security elliptic-curve cryptography. Best of all, DNSCurve has very low overhead and adds virtually no latency to queries.
DNSCurve aims at securing the entire chain down to authoritative servers. However, it only works with authoritative servers that explicitly support the protocol. And unfortunately, DNSCurve hasn't received much adoption yet.
The DNSCrypt protocol 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 link in the chain) against man-in-the-middle attacks. It also provides some confidentiality to DNS queries.
The DNSCrypt daemon acts as a DNS proxy between a regular client, like a DNS cache or an operating system stub resolver, and a DNSCrypt-aware resolver, like OpenDNS.
The daemon is known to work on recent versions of OSX, OpenBSD, NetBSD, Dragonfly BSD, FreeBSD and Linux.
Compile and install it using the standard procedure:
$ ./configure && make -j2
# make install
Replace -j2
with whatever CPU cores you want to use for the
compilation process.
Running make -j2 test
in the src/libnacl
directory is also highly
recommended.
On BSD systems, GNU Make should be installed prior to running the
./configure
script.
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.
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
The proxy will accept incoming requests on 127.0.0.1 and encrypt/decrypt them from/to OpenDNS resolvers.
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>
in order to locally bind a different IP address than 127.0.0.1--local-port=<port>
to change the local port to listen to.--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.--max-active-requests=<count>
to set the maximum number of active requests. The default value is 250.--pid-file=<file>
in order to store the PID number to a file.--user=<user name>
in order to chroot()/drop privileges.
DNSCrypt comes pre-configured for OpenDNS, although the
--resolver-address=<ip>
, --provider-name=<certificate provider FQDN>
and --provider-key=<provider public key>
can be specified in
order to change the default settings.
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, like unbound
, pdns
or
dnscache
.
Both can safely run on the same machine as long as they are listening to different IP addresses or different ports.
If your DNS cache is unbound, all you need is to edit the
unbound.conf
file and add the following lines to the server
section:
forward-zone:
name: "."
forward-addr: 127.0.0.1@40
Then start dnscrypt-proxy
, telling it to use a specific port (40
, in
this example):
# dnscrypt-proxy --local-port=40 --daemonize
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 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-port=<port>
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 encrypts and signs the query and forwards it over TCP
to the resolver.
TCP is slower than UDP, and this workaround should never be used except when bypassing a filter is actually required. Moreover, multiple queries over a single TCP connections aren't supported yet.
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
signing and encrypting them, in order to add the EDNS0 mechanism. By
default, a conservative payload size of 1280 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.
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.