Password-less registration and login

I have been thinking about password-less login lately, and wondering why it's not used more. A couple of days ago I was reading a thread on and came across this blog post by Nick Moore on the subject. It inspired me to post my own thoughts.

I hate having to remember different passwords for different sites. Password managers help, but I often end up using a site's “password reset” feature as a one-time password service. That is: every time I come to login and use the site I reset my password. I suspect I'm not the only one. So how about we make that the default login flow?

The basic idea is that rather than supplying an email and a password when you register or sign in to a site you just supply the email address. The site sends you an email with an embedded one-time password, and visiting that link within N minutes will show a big fat "Confirm Login" or "Confirm Registration" button as appropriate. If the link has been used before, or you wait longer than N minutes before visiting, the page simply says the link is no longer valid and prompt you to go through the process again.

The main benefit to users is that they don't have to remember (or manage) yet another password for yet another site. There's even more benefits for the site maintainers though. Since there's no user-supplied passwords, there's:

  1. No need to maintain a separate password reset functionality: users always go through the regular login procedure.
  2. No need to maintain client-side password strength checks.
  3. Less chance an attacker can simply guess or brute-force a password. (Let's assume we're not up against clairvoyant attackers that can reliably guess UUID-style passwords on the first attempt.)
  4. The registration step has baked-in email verification, so there's no need to maintain a separate path for that either.

There's only really one snag I can think of compared to the traditional login + password approach: We rely on users having convenient access to their email on the system they're logging in from. So this scheme would not be ideal if you're not logging in from your own computer. However, many of us carry around our computers these days; and most mail providers offer web mail, so I don't think this is a major roadblock.

Taken to the extreme this snag could land users in real trouble if they registered with an email address they no longer have access to. With the traditional approach you could luck out and remember the password, and thus be able to login. (Although whether you' d be able to update your email address is another matter: well-behaved sites usually send an email to your old address for confirmation that you actually want to update.) Recommending people register a secondary email address, or a phone number that can receive SMS, could save some support emails in this situation.

I can imagine some people objecting to requiring an email round-trip to complete user registration. However, many sites already require email verification after registration and I don't see this being substantially different. If anything, it's simpler to maintain because there's just one path. And there's nothing stopping you from letting them start using your site straight away and just rely on session state until they've finished registering.

I've been mulling over the idea of password-less registration and login for some time, and for the above reasons I am fairly confident it will work. I don't think there are any glaring security issues compared to the traditional approach; I am indeed confident it is more secure.

Update 2014/01/07: I've received several pieces of feedback on this post. First a friend pointed out in private email:

There's another way around this problem: something like oauth. Sure, it's a fucker to implement correctly, but there are other systems out there that offer similar capabilities (Facebook login, your google ID, Twitter also support authentication) I don't see what benefit this approach offers over those.

Personally I find OAuth, particularly in mobile apps, a terrible user experience. It's just about bearable in browser-based apps, but the way it bounces you back and forth between native mobile apps makes it seem like your phone is having an epileptic fit. My dad would probably throw the device across the room in the belief it was possessed. (And would call me to exorcise it.)

If you've used computers in anger for a while it's easy to get forget all the horrible things we put up with in the name of security, but I think OAuth in particular seriously alienates non-technical users. ("I went to log in, and suddenly I was at a different site! Crazy!") And from the perspective of the site maintainer, if OAuth really is "a fucker to implement correctly" that alone is a fairly strong argument against it in my opinion.

Facebook Login, Google ID and Twitter Auth might be simpler to implement than OAuth. (Disclaimer: I don't know—I've never implemented any of them.) And all have the benefit that they are well known services, and so many people trust them—perhaps more than they trust your site! However they all also suffer the same problem: they require you to have a particular type of account. I agree that if your site integrates heavily with Facebook or Twitter then it would be natural to support Facebook login or Twitter Auth. However if your site offer Google ID and Facebook login but someone visiting your site only has a Twitter account, well; that person probably won't register.

Offering all the possible options for third-party login isn't practical, and as a site maintainer you probably shouldn't try. (In Paradox of Choice Barry Schwartz argues that if you give people too many options it increases their anxiety and the risk that they can't make up their mind—thus walking away with unfinished business.) Finally, as a site maintainer you have to trust not only that those services will stick around in the long term (they probably will!) but that they will continue to support this service. (I seem to recall that Twitter discontinued one authentication mechanism and left a lot of users scrambling for an alternative.)

An email address on the other hand (at least if it's one from a domain you own) follows you and is not relying on any particular company's willingness to continue operating a service. Email has other problems though, as jcs—the "owner" of previously mentioned—pointed out in this comment:

… aaand this is where my enthusiasm for this idea is starting to falter. I will mull this over a while longer before putting the above password-less login flow into use.


In the course of a day I've learnt many reasons why the scheme I proposed, although having some nice properties in theory, is unlikely to be successful in practice. I've rediscovered that I can save a lot of time by—instead of diving into implementation—presenting an idea and asking people smarter than me what's wrong with it. I should do this more.

Posted 5 December 2014 Copyright © Stig Brautaset