|Angristan 7f6a007637||3 years ago|
|LICENSE||3 years ago|
|README.md||3 years ago|
|openvpn-install.sh||3 years ago|
Secure OpenVPN installer for Debian, Ubuntu, CentOS and Arch Linux.
This script will let you setup your own secure VPN server in just a few minutes.
Here is a preview of the installer :
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.
First, get the script and make it executable :
wget https://raw.githubusercontent.com/Angristan/OpenVPN-install/master/openvpn-install.sh chmod +x openvpn-install.sh
Then run it :
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 :
This script is based on the great work of Nyr and its contributors.
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.
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.
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 However I don’t see why you would use an outdated client.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.
The script is made to work on these OS :
(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.
You can get a cheap VPS to run this script for 3€/month at PulseHeberg.
This fork includes the following features :
The script will ask you which DNS resolvers you want to use when connected to the VPN.
Here are the possibilities :
Any other fast, trustable and neutral servers proposition is welcome.
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 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. 😉
OpenVPN uses TLS 1.0 by default, which is nearly 20 years old.
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.
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.
OpenVPN uses an RSA certificate with a 2048 bits key by default.
2048 bits is OK, but both NSA and ANSSI 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.
OpenVPN uses SHA-256 by default.
However SHA-256 is vulnerable to the Length extension attack. OpenVPN supports MD5 and SHA-1, but they’re also vulnerable. This leaves us with SHA-384 and SHA-512. I didn’t find any reason to use SHA-512 over SHA-384, so I’ve chosen SHA-384.
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.
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 from OpenVPN’s documentation says :
The following ciphers are affected, and should no longer be used:
The following ciphers are not affected:
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.
Currently AES is only available in its CBC mode, which is weaker than GCM.
To quote the OpenVPN documentation :
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 weel as ECDH and ECDSA) as soon as OpenVPN 2.4 is released.
For now, these cipher are available in the setup :
Moreover, AES-256 is more vulnerable to Timing attacks.
Thus, the best data channel cipher currently available in OpenVPN is
According to the Hardening page of the OpenVPN wiki, TLS 1.2 is not supported by OpenVPN <2.3.3, so it uses a TLS 1.0 cipher by default, which is unsecure.
The following are TLSv1.2 DHE + RSA choices, requiring a compatible peer running at least OpenVPN 2.3.3:
AES GCM is more secure than AES CBC, and AES 128 is secure enough today. However, the AES-256-GCM cipher uses SHA-384 and the AES-128-GCM one uses SHA-256, however SHA-256 is vulnerable to the Length extension attack. Also, the control channel is a low-bandwidth channel so the ~40% speed difference between AES 128 and 256 bits is nigligible.
Thus, I have chosen
TLS-DHE-RSA-WITH-AES-256-GCM-SHA384 as the control channel cipher.
OpenVPN uses a 2048 bits DH key by default.
2048 bits is OK, but both NSA and ANSSI 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.
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.
As said before, anything lower than SHA-256 is weak, and SHA-256 itself is vulnerable.
Thus I have chosen SHA-384 for the HMAC authentication digest algorithm.
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.
TLS-Auth is not enabled by default by OpenVPN, but it is in this script.
Thanks to the contributors and of course Nyr’s orginal work.