back to article Security? Don't bother until it's needed says RFC

All-or-nothing approaches to security are part of what's making it so hard to achieve acceptable protection, a new RFC suggests. Written by Viktor Dukhovni of Two Sigma, RFC 7435 argues that the way current systems fail is a discouragement to good security. A binary failure – if two peers in a conversation don't have the same …

  1. John Robson Silver badge

    Good idea

    Conditional on a couple of things:

    - Some UI element which tells the user what level of security is being used

    - Some mechanism to define a minimum allowed (still start low and work up, but don't connect if you can't pass some minimum threshold) [NB looks like this is included]

    Particularly if I can tell my machine to always ask if a connection won't exceed "level 3"

  2. Richard Jones 1
    Happy

    I'm Probably Reading This Wrong

    On the face of it this makes total sense. There is less than no point in firing off a secured message if the only person who could ever read it is the sender. On the other hand if you need a secure exchange of information it goes against the grain to send it in clear text when it could have been more securely exchanged. Is this not tantamount to creating a transient virtual VPN on the fly just for the duration of a message exchange.

    At first I was ready to think that it would not apply to me. Then I realised the only times I have ever encrypted anything was to prevent casual reading by other than the recipient. Prior to transmission I agreed the method and applied the method and the agreed decoding mechanism so I guess the exchange followed in a manual form what is being proposed - or have I completely missed the point?

    I also tend to only converse with people who speak the language I use though this does not always avoid communications problems.

  3. Anonymous Coward
    Anonymous Coward

    Is it me, or is the main reason secure transmission fails is because the channel fails to provide the minimum acceptable level? Taking the proposal the other way, why not have a "fail gracefully" system, in which you first try the most secure method you have, then if that doesn't work, fall back step by step to something not so tight but still useable until you're finally forced to drop because you've fallen below the least acceptable?

  4. choleric

    Not new

    This is the way encryption between email servers operates. You can generally set the minimum acceptable level of encryption and you can set your server to only send/receive email if the connection is encrypted if you want to. The big problem is that there are so many email systems that do not or cannot encrypt and if you insist on encryption then you cut yourself off from them and them from you.

    Opportunistic encryption offers a way to include these encrypt-holdouts but at the (huge) expense of guaranteed encryption to hosts that support it. How does an email server easily tell the difference between another server that doesn't support encryption and a compromised connection to a server that in fact does but the mitm is fiddling it so he can read the email?

    This solution just seems to systemically embed the generally ambivalent prevailing social attitude towards genuine privacy and security. People don't value what they have in meatspace and aren't bothered about ensuring it online.

    1. Charles 9

      Re: Not new

      "This solution just seems to systemically embed the generally ambivalent prevailing social attitude towards genuine privacy and security. People don't value what they have in meatspace and aren't bothered about ensuring it online."

      IOW, you hit the meatbag problem, "How do you educate people who don't care yet can threaten you with their imcompetence?"

  5. jake Silver badge

    The real problem is ...

    ... that the vast majority of computer users (including most so-called system administrators!) have absolutely zero concept of personal security, much less digital security.

    Why? Because most people are willfully, intentionally and stubbornly ignorant. They just want to coast along, without thinking, having faith in what they think they know to be "facts"[1].

    No number of RFCs will ever change this.

    [1] "It ain't what you know, it's what you know that just ain't so." -- Sam Clemens

  6. Dan 55 Silver badge
    Thumb Down

    Not the right way to do this

    Hence Dukhovni suggests where suitable (more on “suitable” shortly), the assumption should be reversed: peers assume that their starting point is cleartext, check each others' advertised security capabilities, and incrementally work up to the best security they're both able to support.

    STARTTLS does this and can be easily MITM'd. I used to think adding a STARTTLS request was a fast, good and cheap way of securing e-mail in transit between servers, then I changed my mind because the MITM fiddles with the cleartext capabilities list. In hindsight it's totally obvious it wasn't good.

    1. Steen Larsen

      Re: Not the right way to do this

      > STARTTLS does this and can be easily MITM'd.

      In think one of the points of the RFC is that TLS which can be easily MITM'd is still better than no TLS at all aka clear-text.

      Amongst other, it protects against passive attacks such as the typical mass surveillance.

      1. T. F. M. Reader

        Re: Not the right way to do this

        TLS which can be easily MITM'd is still better than no TLS

        I am not sure. I think I'd prefer to know I am communicating in clear text than to be lulled into a false sense of privacy.

  7. itzman

    It has to come..

    Despite witterings by clueless politicians, routine encryption of nearly all IP packets between nodes with keys known only to those nodes is probably something that with IPV6, will slowly happen in time

    IPV4 was built with resilience in mind. We need a new net built with security in mind.

    What would it matter if the same end user tools ran over a layer than no one had to make a decision to install, but which provided a secure end to end link? As a matter of course?

    Well you would have to move your intel to the end points instead of hoping for a free lunch as a MITM.

    1. Yes Me Silver badge

      Re: It has to come..

      This has nothing to do with IPv6. Really nothing. It has to do with default behaviour built into popular application-layer protocols.

  8. Haro

    End storage

    I suspect that all the carefully encrypted emails get stored in clear text on a Windows machine. Busy executives usually have 5000 inbox emails in Outlook. They may encrypt the whole laptop, but the French police know how to take care of that. :)

  9. Will Godfrey Silver badge

    Hmmm.

    Surely, by the time you know you need security it's already too late.

  10. Anonymous Coward
    Anonymous Coward

    RFC or KFC

    Did that advice come from RFC or was it KFC?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like