Slashdot Mirror
Dutch Government Officially Trusts OpenVPN-NL
First time accepted submitter joost.bijl writes "Yesterday the Dutch government took a step to further improve the adoption of Open Source in its ranks. It has officialy approved a modified version of the open source VPN software OpenVPN for use on the governmental level 'Departementaal Vertrouwelijk' (Restricted). The release is called OpenVPN-NL and is fully open-source and available for use. The software has undergone a security evaluation by the Dutch government's national communications security agency (NLNCSA). The major change is the removal of OpenSSL as the cryptographic core of OpenVPN-NL. Instead, the Dutch government opted to include the smaller, better readable and documented open source library PolarSSL to provide the cryptographic and SSL/TLS functionality. The Dutch IT Security company Fox-IT worked together with both OpenVPN and PolarSSL communities and modified the stock software to support the government evaluation process. In total 8000 lines of code and 4000 lines of documentation were checked in to the OpenVPN trunk."
This is very good news. OpenVPN is probably the easiest secure VPN software I've ever worked with. I've been running it as the link for our multi-site network for over two years now, and it's also the VPN software our road warriors are using. Simple to configure, and damnit but it just works. After years of trying to get all these weird implementations of IPSec to co-operate with each other, OpenVPN is just a marvel, fast and lightweight.
The world's burning. Moped Jesus spotted on I50. Details at 11.
OpenSSL only goes up to TLS1.0, which contains some vulnerabilities. (Note sure if these issue affect OpenVPN). PolarSSL (which is created by a Dutch company, which might be the reason that was chosen) supports up to TLS1.1.
Why they didn't go for the more feature complete and mature GnuTLS would be an interesting question.
http://en.wikipedia.org/wiki/Comparison_of_TLS_Implementations
That's true, though openssl has had the ability to add empty fragments to avoid the chosen plain text attack I suspect you're referring to for many years. What's strange is that the chosen solution (polarSSL) doesn't seem to have support for OCSP which is the main way to quickly revoke bad keys - particularly important in the light of the recent diginotar breach.
When VPN routers were hard to find I set up several OpenVPN links. Over the years most of those networks migrated to other VPN solutions but this one never changed and it always worked. Meanwhile I had to dick with the other solutions all the damn time. When the client with that old OpenVPN link wanted another link I took a good hard look at it. I never had to reconfigure it. I never had to reboot it. It was installed on two HP desktop mini-towers that the client gave to me. And I realized just how good that product was. So I used OpenVPN for the two new links, too. But I upgraded to version 2 and used Centos. That one has been up for two months and everyone is pleased as punch. I'm about to take the old one out of service and install a newer machine running version 2. I'm sure they'll last another ten years.
Holland has made a wise decision to support OpenVPN!
No one ever had to evacuate a city because the solar panels broke!
Yeah, those silly Dutch just don't have a clue... By the way, is the United States still using Windows to control their nuclear power plants???
you don't know what you're talking about. Openvpn was never affected by the "renegotiation bug" as it doesn't use SSL for that component. As it runs over UDP and TCP, it had to come up with its own way of doing that - hence no problem.
That in combination with HMAC authentication makes it basically immune from that issue anyway...
Differences in code between OpenVPN and OpenVPN-NL. (credits: Palatinux) openvpn_nl-v2.1.4-diffpatch.txt
About why the chose to use PolarSSL:
Among the notable differences between OpenVPN and OpenVPN-NL is the cryptographic library. Correct SSL functionality is essential for the protection that OpenVPN offers. OpenSSL is a large and complex library. PolarSSL is a compact and modular library, which is small enough for a fairly in-depth evaluation. Therefore, in the OpenVPN-NL package, it has been chosen to exchange PolarSSL for OpenSSL. This change does not change functionality; the two libraries (OpenSSL and PolarSSL) are mutually compatible.
source: background OpenVPN
But as being said in another comment, someone now working for Fox-IT was involved in PolarSSL. Extra functionality and documentation was added to PolarSSL by Fox-IT according to a comment on a tech-site (tweakers.net) by someone who claims to be the maintainer of PolarSSL.
This seems like a sensible move. It also seems like a major endorsement for OpenVPN. I've always had better experience with OpenVPN than with other VPN solutions, but I have the feeling it hasn't gained much traction. This may be a step in the right direction.
Also, I hadn't heard of PolarSSL, but it sounds worth checking out. OpenSSL has always worked for me, but it is true that the interfaces and documentation aren't the best I've ever seen.
Please correct me if I got my facts wrong.
I think the issue is readability and documentation (and why, that's just what it says!)
Years back I wrote an encryption program in C as an exercise for myself using the OpenSSL libraries partly to learn how to use the APIs. Just a simple wrapper around well-documented APIs, knock it up in no time, right? Wrong!
The documentation was almost unintelligible to anyone who was not an OpenSSL developer or not prepared to study up on the algorithms used and wade through the OpenSSL code base to understand what the APIs did. I doubt that has changed. I found a tutorial giving code snippets in a popular journal that were incorrect and had a crucial error resulting in much hair-pulling. I eventually solved this by pure guesswork and trial-and-error. It would have been much quicker to just cut and paste from someone else's openssl-based encryption program. This all reminded me of those frustrating time-wasting assignments as a student where lecturers forgot to tell us that it just wouldn't work or be solvable without secret Factor X and someone in the class had to discover this sideways from a tutor.
Time OpenSSL grew up and stopped living in arcane land. It needs decent docs. There is an O'Reilly book on it - maybe that'd help next time.