DNSSEC - what is it and how it works
Every time a device connected to the internet wants to communicate with another device (e.g. your computer and A/I servers) it must do so by contacting said device on its IP address. It would be unconvenient to do all of that manually and remember IP addresses, thus the DNS has been invented to translate names (e.g. www.autistici.org) into IP addresses (e.g. 18.104.22.168).
DNS itself doesn't provide any guarantee over authenticity of its replies. This means that anybody controlling your traffic is capable of changing the IP addresses your computer sees. This scenario is exploited for example for very simple censorship: if the site is censored then your ISP DNS servers will return no IP addresses instead of the real ones.
The DNSSEC protocol has been introduced to provide an additional layer of security by making sure that replies can be authenticated. DNSSEC is not perfect (see for example the problem of zone enumeration and criticism by D. J. Bernstein) however we think that it is important to provide users and additional validated channel.
How do I use it?
One of the simplest ways to use DNSSEC is to use a public DNSSEC-validating DNS (such as google public dns). However this solution assumes that the communication between your device and the DNS server is to be trusted.
Instead, you can run locally a validating DNS server that will do the validation. There are pre-packaged solutions to ease this task such as dnssec-trigger oriented to installation on roaming laptops.
Is it working?
You can verify that DNSSEC is working properly by assessing whether it is working or not. Specifically, there are websites that are broken on purpose (e.g. dnssec-failed.org) and your device won't be able to use them. In particular your device won't be able to know the website's IP address. The command below can be used to verify using the terminal (if for example you are using a local resolver as explained above). Notice the command resuts no results (status: SERVFAIL):
$ dig +dnssec +noauth +noquestion +nocmd +nostats dnssec-failed.org @localhost ;; Got answer: ;; ->>HEADER<<- opcode: QUERY, status: SERVFAIL, id: 49462 ;; flags: qr rd ra; QUERY: 1, ANSWER: 0, AUTHORITY: 0, ADDITIONAL: 1 ;; OPT PSEUDOSECTION: ; EDNS: version: 0, flags: do; udp: 1024On a valid domain however be the valid answer flag "ad" will be set by the local resolver:
$ dig +dnssec +noauth +noquestion +nocmd +nostats autistici.org @localhost ;; Got answer: ;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 63802 ;; flags: qr rd ra ad; QUERY: 1, ANSWER: 4, AUTHORITY: 6, ADDITIONAL: 1 ;; OPT PSEUDOSECTION: ; EDNS: version: 0, flags: do; udp: 1024 ;; ANSWER SECTION: autistici.org. 9382 IN A 22.214.171.124 autistici.org. 9382 IN A 126.96.36.199 autistici.org. 9382 IN A 188.8.131.52 autistici.org. 9382 IN RRSIG A 7 2 9600 20140110201142 20131211201142 24242 autistici.org. VMeyEaa3iiCwlDwhch6erY0ScJkdli9P+xzoRP27Ww7/hGea3rwo4hpX 3M8HFwDIkE3aQQLDwvSvlFcB0XzvKdrbOgKXRW4nvaAvhEO5E8eKYmgb oMphm2RoU1+I6r+WWpKpaE/o6bOzDgZ43j1KkjOuEODA/3vbN7G7gIg8 JDw=
How to validate using the command line?
If you are using a local validating DNS, all answers directed to that server will be already validated. In cases where this isn't possible you can still manually validate a request (in a slighly convoluted way):
The first step is to obtain the root "key", e.g.
dig +nocomments +nostats +nocmd +noquestion -t dnskey . > trusted-root.keythen use this key as a trusted key used to validate answers:
dig +sigchase +trusted-key=trusted-root.key www.autistici.org A
The hierarchial nature of DNS is also reflected in the DNSSEC deployment, this means that each entity in the DNS tree can autonomously publish and generate its keys. Compare this to the existing CA infrastracture for https certificates with a few valid entities that are globally trusted. With DNSSEC it is possible to freely change the keys by simply publish them to your parent zone. (e.g. autistici.org publishes its keys towards org.)
DNS supports other data types other than IP addresses, e.g. you could publish SSL certificate information for a given domain. The DANE protocol aims at specifying the details of TLS certificate fingerprints over DNS and bypassing the root CA system altogether. Our CA verification page uses this protocol to establish a chain of trust via DNS.