**You have to enable the TUN module otherwise OpenVPN won't work.** Ask your host if you don't know how to do it. If the TUN module is not enabled, the script will warn you and exit.
You can get a cheap VPS to run this script for $2.50/month worldwide at [Vultr](https://goo.gl/Xyd1Sc) or 3€/month for unlimited bandwidth in France at [PulseHeberg](https://goo.gl/76yqW5).
The first time you run it, you'll have to follow the assistant and answer a few questions to setup your VPN server.
When OpenVPN is installed, you can run the script again, and you will get the choice to :
- Add a client
- Remove a client
- Uninstall OpenVPN
## The fork
This script is based on the great work of [Nyr and its contributors](https://github.com/Nyr/openvpn-install).
I made it because I wanted to have a more secured OpenVPN out-of-the-box. It works like the original script, but is more focused on privacy and espicially better encryption. Nyr's original script uses mainly default parameters regarding encryption, and some of them are unsecure. See [#encryption](#encryption).
Also, Nyr and myself clearly have not the same point of view regarding this script, that's why it's a fork.
The only drawback is that you need to use a recent version of OpenVPN, because some parameters that requires TLS 1.2 are only availble since OpenVPN 2.3.3. Therefore I restrain the compatibility of this script to a few but widely used GNU/Linux distributions, to get a recent version of OpenVPN from trusted third-party repositories, if needed. That is not a complete drawback tough, because it means that you can have the latest version with all the new features and security fixes. See [compatibilty](#compatibility).
On the client-side, it's less problematic, but if you want to use an OpenVPN server installed with this script with an old client (\<2.3.3), it won't work. However I don't see why you would use an outdated client.
**TL;DR**, this script is relatively secure, and you can just press enter in the setup.
**[A Pull Request](https://github.com/Angristan/OpenVPN-install/pull/96) is currently being worked on to implement the latest OpenVPN 2.4 features.**
## Compatibility
The script is made to work on these OS and architectures :
- **Debian 7** (i386, amd64)
- **Debian 8** (i386, amd64)
- **Debian 9** (i386, amd64, armhf, arm64)
- **Ubuntu 12.04 LTS** (i386, amd64)
- **Ubuntu 14.04 LTS** (i386, amd64)
- **Ubuntu 16.04 LTS** (i386, amd64)
- **Ubuntu 16.10** (i386, amd64, armhf, arm64)
- **Ubuntu 17.04** (i386, amd64, armhf, arm64)
- **Fedora 25** (amd64)
- **Fedora 26** (amd64)
- **Fedora 27** (amd64)
- **CentOS 6** (i386, amd64)
- **CentOS 7** (i386, amd64, arm64)
- **Arch Linux** (i686, amd64, arm64)
(It should also work on Debian unstable/testing and Ubuntu beta).
If your're using an Ubuntu version that is not supported by the script, be aware that it's not supported by Ubuntu either, thus it's insecure.
## Features
This fork includes the following features :
- Every feature of the [original script](https://github.com/Nyr/openvpn-install)
- Better encryption, see below
- Better DNS resolvers, see below
- Choice between TCP and UDP (UDP is still recommended)
- Run server in [unprivileged mode](https://github.com/Angristan/OpenVPN-install/blob/master/openvpn-install.sh#L426), reducing risks to the system
- [Block DNS leak on Windows 10](https://community.openvpn.net/openvpn/ticket/605)
- No comp-lzo, as [compression is a vector for oracle attacks, e.g. CRIME or BREACH](https://github.com/BetterCrypto/Applied-Crypto-Hardening/pull/91#issuecomment-75388575)
- [Arch Linux support](https://github.com/Angristan/OpenVPN-install/pull/2)
- Up-to-date OpenVPN thanks to [EPEL](http://fedoraproject.org/wiki/EPEL) for CentOS and [swupdate.openvpn.net](https://community.openvpn.net/openvpn/wiki/OpenvpnSoftwareRepos) for Ubuntu and Debian. These are third-party yet trusted repositories.
- Randomized certificate name
- Other improvements !
## DNS
The script will ask you which DNS resolvers you want to use when connected to the VPN.
Here are the possibilities :
- Current system resolvers, those that are in `/etc/resolv.conf`
- [Quad9](https://www.quad9.net), recommended, security and privacy oriented, fast worldwide (Anycast servers)
- [FDN's DNS Servers](http://www.fdn.fr/actions/dns/), recommended if you're in western europe (France)
- [DNS.WATCH DNS Servers](https://dns.watch/index), recommended if you're in western europe (Germany)
- [OpenDNS](https://en.wikipedia.org/wiki/OpenDNS), not recommened but fast wordlwide (Anycast servers)
- [Google Public DNS](https://en.wikipedia.org/wiki/Google_Public_DNS), not recommended, but fast worldwide (Anycast servers)
- [Yandex Basic DNS](https://dns.yandex.com/), not recommended, but fast in Russia
- [AdGuard DNS](https://github.com/AdguardTeam/AdguardDNS), located in Russia, blocks ads and trackers
- Soon : local resolver :D
Any other fast, trustable and neutral servers proposition is welcome.
## Encryption
The main reason why I made this fork is to improve the encryption. Indeed, OpenVPN's default parameters are weak (and that's what [Nyr's script](https://github.com/Nyr/openvpn-install) uses).
I want to justify all my choices regarding the encryption settings I have chosen, to prove that I'm not some random noob as some may think. 😉
However I'm far from a crypto expert, so don't hesitate to doubt what I say (I put links to my sources anyway), and to open an issue to correct me.
OpenVPN 2.4 will be a great update on the encryption part, because we'll be able to use elliptic curves, so ECDSA and ECDH (as well for the control channel), and AES GCM. They are faster and more secure. I will, of course, update the script when it will be available.
**Note:** With OpenVPN's default parameters, you have a relatively weak encryption. Nonetheless, your trafic is still encrypted, so unless you're under surveillance, probably no one will try to decrypt it. Yet it's not a reason to use old and weak algorithm when there are much better ones available. 😉
### TLS version
OpenVPN uses TLS 1.0 by default, which is nearly [20 years old](https://en.wikipedia.org/wiki/Transport_Layer_Security#TLS_1.0).
With `tls-version-min 1.2` we use at least TLS 1.2, which the best protocol available currently. I could have used `tls-version-min highest` but this does not ensure we use TLS 1.2 which is the only secure protocol available.
[OpenVPN documentation for tls-version-min](https://community.openvpn.net/openvpn/wiki/Openvpn23ManPage#lbAK)
TLS 1.2 is only supported since OpenVPN 2.3.3. This is one of the reasons of the script uses third-party repositories, because some distributions have an older version of OpenVPN.
### Certificate
#### Key
OpenVPN uses an RSA certificate with a 2048 bits key [by default](https://github.com/OpenVPN/easy-rsa/blob/master/easyrsa3/vars.example#L97).
2048 bits is OK, but both [NSA](https://cryptome.org/2016/01/CNSA-Suite-and-Quantum-Computing-FAQ.pdf) and [ANSSI](https://www.ssi.gouv.fr/uploads/2015/01/RGS_v-2-0_B1.pdf) recommend at least a 3072 bits for a future-proof key. As the size of the key will have an impact on speed, I leave the choice to use 2048, 3072 or 4096 bits RSA key. 4096 bits is what's most used and recommened today, but 3072 bits is still good.
In OpenVPN 2.4, we will be able to use an ECDSA certificate. This algorithm uses elliptic curves instead of prime numbers' factorization for a reduced key size and calculation time, thus it's faster and more secure.
It also supports SHA1 and MD5, which are unsafe, and all the SHA2 family. I didn't find any reason to use something other than SHA-256 in the SHA2 group, so the script still uses the default hash algorithm.
### Data channel's cipher
By default, OpenVPN uses `BF-CBC` as the data channel cipher. Blowfish is an old (1993) an weak alogorithm. What's *funny* is that even the official OpenVPN documentation admits it.
>The default is BF-CBC, an abbreviation for Blowfish in Cipher Block Chaining mode.
Using BF-CBC is no longer recommended, because of its 64-bit block size. This small block size allows attacks based on collisions, as demonstrated by SWEET32. See https://community.openvpn.net/openvpn/wiki/SWEET32 for details.
>Security researchers at INRIA published an attack on 64-bit block ciphers, such as 3DES and Blowfish. They show that they are able to recover plaintext when the same data is sent often enough, and show how they can use cross-site scripting vulnerabilities to send data of interest often enough. This works over HTTPS, but also works for HTTP-over-OpenVPN. See https://sweet32.info/ for a much better and more elaborate explanation.
> OpenVPN's default cipher, BF-CBC, is affected by this attack.
>Blowfish's use of a 64-bit block size (as opposed to e.g. AES's 128-bit block size) makes it vulnerable to birthday attacks, particularly in contexts like HTTPS. In 2016, the SWEET32 attack demonstrated how to leverage birthday attacks to perform plaintext recovery (i.e. decrypting ciphertext) against ciphers with a 64-bit block size such as Blowfish.[9]
>A reduced-round variant of Blowfish is known to be susceptible to known-plaintext attacks on reflectively weak keys. Blowfish implementations use 16 rounds of encryption, and are not susceptible to this attack. Blowfish users are encouraged by Bruce Schneier, Blowfish's creator, to use the more modern and computationally efficient alternative Twofish. He is quoted in 2007 as saying:
>"At this point, though, I'm amazed it's still being used. If people ask, I recommend Twofish instead."
The [SWEET32 vulnerability page](https://community.openvpn.net/openvpn/wiki/SWEET32) from OpenVPN's documentation says :
>The following ciphers are affected, and should no longer be used:
- BF-*
- DES* (including 3DES variants)
>The following ciphers are *not* affected:
- AES-*
- CAMELLIA-*
- SEED-*
Indeed, AES is today's standard. It's the fastest and more secure cipher available today. [SEED](https://en.wikipedia.org/wiki/SEED) and [Camellia](https://en.wikipedia.org/wiki/Camellia_(cipher)) are not vulnerable to date but are slower than AES and relatively less trusted.
As they have not any proven vulnerabilities, I decided to give the user the choice to use them, though I don't see any particular reason to this day to use it. Maybe someday if AES happens to be broken. Here is an exemple about [why Camellia is good, but AES is better and should be used](http://crypto.stackexchange.com/questions/476/why-does-nobody-use-or-break-the-camellia-cipher/477#477).
Currently AES is only available in its CBC mode, which is weaker than GCM.
To quote the [OpenVPN documentation](https://community.openvpn.net/openvpn/wiki/SWEET32) :
>Of the currently supported ciphers, OpenVPN currently recommends using AES-256-CBC or AES-128-CBC. OpenVPN 2.4 and newer will also support GCM. For 2.4+, we recommend using AES-256-GCM or AES-128-GCM.
Of course I will update the script to add AES-GCM mode (as well as ECDH and ECDSA) as soon as OpenVPN 2.4 is released.
For now, these cipher are available in the setup :
- AES-128-CBC
- AES-192-CBC
- AES-256-CBC
- CAMELLIA-128-CBC
- CAMELLIA-192-CBC
- CAMELLIA-256-CBC
- SEED-CBC
AES-256 is 40% slower than AES-128, and there isn't any real reason to use a 256 bits key over a 128 bits key with AES. (Source : [[1]](http://security.stackexchange.com/questions/14068/why-most-people-use-256-bit-encryption-instead-of-128-bit),[[2]](http://security.stackexchange.com/questions/6141/amount-of-simple-operations-that-is-safely-out-of-reach-for-all-humanity/6149#6149)).
Moreover, AES-256 is more vulnerable to [Timing attacks](https://en.wikipedia.org/wiki/Timing_attack).
Thus, the best data channel cipher currently available in OpenVPN is `AES-128-CBC`.
### Control channel's cipher
According to the [Hardening](https://community.openvpn.net/openvpn/wiki/Hardening#Useof--tls-cipher) page of the OpenVPN wiki, TLS 1.2 is not supported by OpenVPN <2.3.3,soitusesaTLS1.0cipherbydefault,whichisunsecure.
> The following are TLSv1.2 DHE + RSA choices, requiring a compatible peer running at least OpenVPN 2.3.3:
- TLS-DHE-RSA-WITH-AES-256-GCM-SHA384
- TLS-DHE-RSA-WITH-AES-256-CBC-SHA256
- TLS-DHE-RSA-WITH-AES-128-GCM-SHA256
- TLS-DHE-RSA-WITH-AES-128-CBC-SHA256
AES GCM is more secure than AES CBC, and AES 128 is secure enough today. I didn't find any security difference between SHA-256 and SHA-384 so we're going to use SHA-256.
Thus, I have chosen `TLS-DHE-RSA-WITH-AES-128-GCM-SHA256` as the control channel cipher.
### Diffie-Hellman key
OpenVPN uses a 2048 bits DH key [by default](https://github.com/OpenVPN/easy-rsa/blob/master/easyrsa3/vars.example#L97).
2048 bits is OK, but both [NSA](https://cryptome.org/2016/01/CNSA-Suite-and-Quantum-Computing-FAQ.pdf) and [ANSSI](https://www.ssi.gouv.fr/uploads/2015/01/RGS_v-2-0_B1.pdf) recommend at least a 3072 bits for a future-proof key. Like RSA, the size of the key will have an impact on speed, I leave the choice to use a 2048, 3072 or 4096 bits key. 4096 bits is what's most used and recommened today, but 3072 bits is still good.
In OpenVPN 2.4, we will be able to use ECDH key. It uses elliptic curves instead of prime numbers' factorization for a reduced key size and calculation time, thus it's faster and more secure.
### HMAC authentication algorithm
To quote the OpenVPN wiki :
>Authenticate packets with HMAC using message digest algorithm alg. (The default is SHA1 ). HMAC is a commonly used message authentication algorithm (MAC) that uses a data string, a secure hash algorithm, and a key, to produce a digital signature.
OpenVPN's usage of HMAC is to first encrypt a packet, then HMAC the resulting ciphertext.
SHA-1 is not safe anymore, so I use SHA-256 which is safe and widely used.
### TLS-Auth
>The --tls-auth option uses a static pre-shared key (PSK) that must be generated in advance and shared among all peers. This features adds "extra protection" to the TLS channel by requiring that incoming packets have a valid signature generated using the PSK key. If this key is ever changed, it must be changed on all peers at the same time (there is no support for rollover.)
>The primary benefit is that an unauthenticated client cannot cause the same CPU/crypto load against a server as the junk traffic can be dropped much sooner. This can aid in mitigating denial-of-service attempts.
>This feature by itself does not improve the TLS auth in any way, although it offers a 2nd line of defense if a future flaw is discovered in a particular TLS cipher-suite or implementation (such as CVE-2014-0160, Heartbleed, where the tls-auth key provided protection against attackers who did not have a copy). However, it offers no protection at all in the event of a complete cryptographic break that can allow decryption of a cipher-suite's traffic.
