Today we announce that we have completely removed all traces of disks being used by our VPN infrastructure!

  • eatham 🇦🇺 ( eatham@aussie.zone ) 
    link
    fedilink
    English
    arrow-up
    105
    ·
    3 years ago

    Full article:

    We have successfully completed our migration to RAM-only VPN infrastructure

    20 September 2023 NEWS SYSTEM TRANSPARENCY

    Today we announce that we have completely removed all traces of disks being used by our VPN infrastructure!

    In early 2022 we announced the beginning of our migration to using diskless infrastructure with our bootloader known as “stboot”. Completing the transition to diskless infrastructure

    Our VPN infrastructure has since been audited with this configuration twice (2023, 2022), and all future audits of our VPN servers will focus solely on RAM-only deployments.

    All of our VPN servers continue to use our custom and extensively slimmed down Linux kernel, where we follow the mainline branch of kernel development. This has allowed us to pull in the latest version so that we can stay up to date with new features and performance improvements, as well as tune and completely remove unnecessary bloat in the kernel.

    The result is that the operating system that we boot, prior to being deployed weighs in at just over 200MB. When servers are rebooted or provisioned for the first time, we can be safe in the knowledge that we get a freshly built kernel, no traces of any log files, and a fully patched OS.

        • Funny thing is I started using Surfshark just before they started all the YouTube sponsorships. Them doing so many sponsorships actually made me trust them less somehow, if that makes sense.

          Mullvad “appears” to be more trustworthy but maybe they are just better at marketing that image. They still cost twice as much as Surfshark.

      • PeachMan ( PeachMan@lemmy.one ) 
        link
        fedilink
        arrow-up
        28
        ·
        3 years ago

        You don’t use Mullvad for their performance, you use them for their insanely paranoid security and privacy practices.

        And for the record, I was never impressed with Surfshark speeds. I dropped them when they bundled a virus scanner into their VPN client, that’s sketchy as hell. I don’t want my VPN provider scanning my files.

      • Onii-Chan ( Onii-Chan@kbin.social ) 
        link
        fedilink
        arrow-up
        12
        ·
        3 years ago

        I never had any real issues with speed using Surfshark, the reason I made the switch was largely about trust. As another user said, as soon as I saw Surfshark start their YouTube advertising spree, and start to bloat their client with unnecessary features, I started looking for alternatives.

        I’m iffy about any VPN company that uses YouTuber marketing as it is, and while my threat model isn’t overly paranoid, I believe the VPN company someone chooses to use should have paranoid business practices. After I saw the news on Mullvad’s raid, the authorities subsequently finding nothing, and the fact that a user’s account is merely a string of numbers, I decided it was the VPN for me.

      • lemmyingly ( lemmyingly@lemm.ee ) 
        link
        fedilink
        arrow-up
        5
        ·
        3 years ago

        Mullvad compared to PIA, Google annoys me less with recaptures. I know it doesn’t answer your question but thought I’d throw my 2 cents in since PIA was quite a popular choice with their YouTube sponsor slots and cheap prices

  • mnmalst ( mnmalst@lemmy.zip ) 
    link
    fedilink
    arrow-up
    34
    ·
    3 years ago

    I find the “Mullvad VPN scratch cards” interesting. If a store near you has them you could buy one and be totally anonymous. What I find a bit odd is that you can buy them on amazon as well but sold directly by mullvad. Doesn’t that defeat the purpose? The idea of the card is a decoupling of your real identity from the vpn user but when you buy the card in their store doesn’t it negate that?

    I am probably just missing something here. Does anyone have more insight?

    • Leraje ( leraje@lemmy.blahaj.zone ) OP
      link
      fedilink
      English
      arrow-up
      58
      ·
      3 years ago

      The code on the card is covered so Amazon might know you use Mullvad but they have no way of knowing what your account is.

      Mullvad know your acct but they have no way of knowing how it is you paid other than maybe it being a scratchcard which they don’t track anyway.

      • mnmalst ( mnmalst@lemmy.zip ) 
        link
        fedilink
        arrow-up
        3
        ·
        3 years ago

        I am not talking about amazon knowing it. Amazon offers shops for businesses, where a business directly sells goods to their customers using amazon as a transaction platform. Those shops send the goods directly to their customers (Sometimes it comes from an amazon warehouse as well tho). If the first case is true, mullvad would send me the card directly, so they would know I bought it, which makes the card obsolete in my view.

        But maybe they don’t send it themself and the cards are all just sitting in a big warehouse. Either way, to me it’s not 100% a given that they couldn’t at least in theory know who bought it.

        I am just playing devils advocate here btw, I am not really concerned about it.

        • You are buying access to a VPN not a nuclear warhead for the black market. The link between buying a VPN card and the code used in that card to link to said vpn activity which is also pretty well protected on Mullvad is not easily discoverable. Seems like a pretty reasonable privacy gap to me.

        • Leraje ( leraje@lemmy.blahaj.zone ) OP
          link
          fedilink
          English
          arrow-up
          12
          ·
          edit-2
          3 years ago

          If you’re a Mullvad customer then they already know your IP and from that they could identify you pretty easily. But that’s true of all VPN providers, but they claim they don’t log and I seem to recall them saying they don’t keep a record of scratch off card numbers (why would they?). Either way you have to trust them and based on the fact they’re totally open I do.

        • Leraje ( leraje@lemmy.blahaj.zone ) OP
          link
          fedilink
          English
          arrow-up
          5
          ·
          3 years ago

          All those things you listed simply confirm that a particular person bought a Mullvad scratch card. There’s literally no way to associate that data with a particular Mullvad account. To do that they’d have to have a record of the card batch number and somehow have accessed the code underneath the scratch-off panel and then find a way to match those numbers against your Mullvad acct.

        • PeachMan ( PeachMan@lemmy.one ) 
          link
          fedilink
          arrow-up
          2
          ·
          3 years ago

          You’re not wrong, BUT that’s why Mullvad offers other forms of anonymous payment, the flexibility lets you be as paranoid as YOU want to be. You can pay in Bitcoin, or you can literally mail them an envelope of cash with no return address. Amazon scratch cards are just the most convenient option, and as always, you trade security for convenience.

    • Simran ( Simran@lemm.ee ) 
      link
      fedilink
      arrow-up
      3
      ·
      3 years ago

      Well the biggest selling point of VPNs is easier piracy not privacy. Most VPN customers just want to protect themselves from anyone watching their downloading habits. Yeah technically there would be a trail but no one is going to follow it to catch someone downloading inception.

  • From what I read in the article, there is still one part of the boot sequence that does require some sort of storage: the part where the bootloader fetches the network boot image and verifies it against the checksum signature. But I think that can be performed by booting from a pendrive and then removing it. The problem will come if law enforcement gets a hold of said pendrive…

    • Leraje ( leraje@lemmy.blahaj.zone ) OP
      link
      fedilink
      English
      arrow-up
      61
      ·
      3 years ago

      Didn’t really have a choice:

      …Regrettably individuals have frequently used this feature to host undesirable content and malicious services from ports that are forwarded from our VPN servers. This has led to law enforcement contacting us, our IPs getting blacklisted, and hosting providers cancelling us.

      Blog post

      Big issue there is hosting providers cancelling them. Can’t operate a business without that.

    • Agreed. Seems like they were in a super tough spot with that and kind of had to drop it. All the sudden they seem to be doing some new cool stuff to try to keep their edge which I really appreciate / respect. That being said, I’ve dumped them and switched to a service that still port forwards as it gives me better torrenting throughput. Sorry Mullvad.

    • viking ( viking@infosec.pub ) 
      link
      fedilink
      arrow-up
      36
      ·
      3 years ago

      Not at all. Of course their operating system has to be booted from some kind of solid state disk, but all actual operations are carried out in the RAM, meaning that nothing is ever written to the disk. Since the RAM is periodically overwritten and doesn’t hold any data in case of a reboot or power failure, they are de-facto not logging a single thing.

      • 4am ( 4am@lemm.ee ) 
        link
        fedilink
        English
        arrow-up
        9
        ·
        3 years ago

        I believe the edge devices would use netboot, so they load the kernel and user land into RAM over a network and have no disk in them at all.

        Here’s hoping that image stays clean😉

        • viking ( viking@infosec.pub ) 
          link
          fedilink
          arrow-up
          1
          ·
          3 years ago

          Right, I mean that the boot image is stored somewhere in a solid state. Whether that’s on a centralized server or locally, no clue. But makes sense to load it remotely, easier to maintain proper versioning that way and ensure cross-compatibility.

    • 👁️👄👁️ ( mojo@lemm.ee ) 
      link
      fedilink
      English
      arrow-up
      2
      ·
      3 years ago

      It’d turn the servers off obviously, but that’d be true if it was on disk or not. The source code is stored elsewhere probably on multiple data servers they access with git.

    • Nollij ( Nollij@sopuli.xyz ) 
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 years ago

      It’s a valid question, even if your scenario isn’t plausible. The very point is that all data is ephemeral - there is no “data at rest” to be compromised. But the problem is that this data is very, very important. It would include (among other things) account information. If all of the servers power off simultaneously (for whatever reason), then yes, it would likely destroy them. More likely is a software fault that causes each system to crash, or lose/corrupt that data.

      But there are ways around this, too. I have no idea which (if any) of these they are doing, just that these are options. They already probably sync data among running servers, it will just now be done exclusively in RAM. They can have “seed” distributed servers, running an entirely different codebase, simply to house this data. They would also be diskless, but mostly unconnected to the standard operational servers. From an architecture and design standpoint, these would work a lot like disks.

      Distributed is also a key word - it wouldn’t be a single server, rack, or even datacenter that would need to collapse. It would be to be all of them, or at least sever their connections to each other.

      (Side note: Going diskless addresses concerns about data security for data at rest. It does nothing about data in motion)

      TL;DR: Theoretically yes, but it would take a lot more than that.

    • Kissaki ( Kissaki@feddit.de ) 
      link
      fedilink
      English
      arrow-up
      11
      ·
      3 years ago

      It’s unlikely to have any noticeable impact. This is more about verifiably and categorically not having any traces of logging or cached state.

      Both caching and logging should be independent of the direct usage performance anyway. And service startup happens only once - not during its usage.