June 17, 2015

Falling Into the Encryption Trap

This is a difficult discussion for me. It borders on embarrassing, because I'm forced to admit that I was unable to foresee some of the ramifications of encryption-related polices I've been promoting for many years.

I could make the excuse that I did not anticipate an onrush of largely hyperbolic paranoia as has been triggered post-Snowden, but I should have realized that some sort of similar event would be adequate to trigger similar issues.

My concern started taking shape about a month and a half ago when discussions were bouncing around the Net regarding Mozilla's supposed intention to (at some point in the future) refuse to allow Firefox browser connections to sites not running SSL/TLS, or at least restricting "features" available to them. Mozilla's actual stance on this is currently not entirely clear, but it appears that their longer-term plans at least are moving in this direction.

Two days ago when I posted When Google Thinks They're Your Mommy, regarding the Chrome browser's refusal to connect to a major corporate site that other browsers would connect to, I triggered a mass of users sending me similar stories about "encryption enforcement" complications -- and not just about Chrome.

There were people complaining about Chrome blocking sites, Firefox blocking sites, new versions of browsers preventing users from accessing network-connected home devices -- that were difficult or impossible to replace and only ran on local networks. On and on. I had let loose an unexpected avalanche, about an issue I incorrectly thought was only now beginning to gradually affect larger numbers of users.

And last night I had a nightmare.

I saw two parents desperately trying to access a misconfigured medical-related site for their sick child, being blocked by Chrome "for their own protection" -- and then trying to install another browser in panic after being informed by a Google help page that Chrome wouldn't help them, and that using another browser was their only alternative.

This is what comes of reading my own blog posts sometimes -- waking up in cold sweat from a very dark dream.

What's happened of course is that post-Snowden there's been a mantra that we're all being spied on all the time, and not only should all Internet connections be encrypted, but users should not be permitted -- no matter what the situation -- to access sites that are not encrypted to the standard that the crypto-gurus feel is adequate, even when the situation is triggered by temporary misconfiguration rather than purposeful configuration decisions at a site.

The argument is that man-in-the-middle attacks are so powerful and so pervasive (the former can certainly be true, the latter is definitely arguable) that even someone viewing kitten videos must use encryption -- if for no other reason than to protect them from some evil entity injecting an exploit into their weak connection.

Obviously if you're Google and continuously transferring gazookabytes of user data between datacenters, you're a big target and you want those circuits to be as rigorously encrypted as is practicable.

The reality though is that the overwhelmingly vast majority of user system exploits aren't based on subversion of the connections at all, but rather on endpoint attacks -- usually tied to phishing and other "social engineering" techniques when available multiple factor authentication systems have not been deployed.

Like I said, I've spent many years promoting the concept of universal, opportunistic Internet encryption.

But in some of the attitudes I see being expressed now about "forced" encryption regimes -- even browsers blocking out fully-informed users who would choose to forgo secure connections in critical situations -- there's a sense of what I might call "crypto-fascism" of a kind.

And that worries me. That's the stuff of nightmares.

It's one thing for a site to specifically and clearly indicate that it will only accept secure connections of a particular class and quality, proclaiming that it feels such restrictions are absolutely necessary in their context.

It's something else entirely though for a browser to unilaterally declare a site's security to be unacceptably weak (perhaps by choice or often by misconfiguration -- both of which we can agree need to be fixed) to the extent that the browser absolutely refuses to allow the user to connect, regardless of how crucial the situation and irrespective of the fully-informed expressed will of the user to connect in any case.

Some encryption and Web standards experts might assert that this is simply a situation where a rather technically fascistic attitude is necessary to protect users overall, even if individual users in some circumstances might be horribly injured in the process. I've already had someone quote Spock to me on this one ("The needs of the many outweigh the needs of the few.")

Leaving aside for the moment that "the few" at Internet scale could easily still be many millions of warm bodies, I don't buy Spock's supposed logic in the context under discussion here.

Yes, we want to encourage encryption -- strong encryption -- on the Net whenever possible and practicable. Yes, we want to pressure sites to fix misconfigured servers and not purposely use weak crypto.

But NO, we must not permit technologists (including me) to deploy Web browsers (that together represent a primary means of accessing the Internet), that on a "security policy" basis alone prevent users from accessing legal sites that are not specifically configured to always require strongly encrypted connections, when those users are informed of the risks and have specifically chosen to proceed.

Anything less is arrogantly treating all users like children incapable of taking the responsibility for their own decisions.

And that would be a terrible precedent indeed for the future of the Internet.


Posted by Lauren at June 17, 2015 12:08 PM | Permalink
Twitter: @laurenweinstein
Google+: Lauren Weinstein