Hacker News Reader

Comment by scrollaway | Hacker News Reader

You don't need a phone at all with TOTP. Just a TOTP client, which could even run on the same machine.

Fun story about this: When Blizzard introduced TOTP, I wrote a Blizzard Auth library and client so that I could use their 2fa on my desktop (https://github.com/jleclanche/python-bna). Why? Because I was sharing my Blizzard account with someone else, but still wanted to use 2fa. And people kept telling me "2fa must be on a different machine otherwise it's not secure", so I was using 2fa wrong and shouldn't use it at all.

Well that's bollocks. Same-machine 2fa is more secure than no 2fa at all. There's another comment below that talks about "perfect being the enemy of good" and that's exactly it.

So we absolutely should start promoting the fact that you can use TOTP on desktop, even if it's less secure. A lot of users don't use 2fa because they either don't have a phone, or don't have their phone constantly with them.

View on HN

My point is someone should have written a TOTP client for your particular platform. Also different services use different TOTP systems. I am not well-versed in the matter. For example Steam uses its own app, many apps use google authenticator, microsoft uses its own, others still use authy. AFAIK they aren't interchangeable. I can't be bothered installing all these apps, each of them has their own nefarious agenda to hoover data off my phone.

SMS just works. (except when it doesn't)

TOTP is a standard. It just works, and it always does, even if you don't have internet on your phone. Even if you don't have reception. Even if you're out of credit. Even if you're out of the country. Even if your telecom provider hates you. Authy is TOTP. Google Authenticator is TOTP. They are all interchangeable. You can use FreeOTP if you don't want to deal with those: https://fedorahosted.org/freeotp/

Steam is SMS, and I heard they now have a shitty custom 2fa but what you're describing is services that don't use an established standard, which is compatible with hundreds of established clients. SMS isn't merely nonstandard (and now deprecated), it's more expensive for both the user (requires an active sim card) and the server (requires sending SMS to numbers all over the world).

The argument I've heard for why steam requires a custom 2fa is to do with trading on the marketplace [1]

The custom component makes sure that marketplace activity is valid (somewhat like the custom banking dongles you get that require both your pin and some transaction identifier).

They'd be better off allowing the option of standard TOTP auth for login, so users can use a standard authenticator app, and then layer on a custom app for trading if needed.

1: http://steamcommunity.com/discussions/forum/0/49463187366895...

Right their needs don't really justify a custom 2FA at all. Security through obscurity isn't all that much of an advantage and bundling the trade approvals and 2FA into the same app is maybe nicer for some users, but technically unnecessary. If anything it just scans as Valve's traditional NIH syndrome being exhibited here.

Can you elaborate? I don't use the marketplace; how does the custom component achieve any of that?

Detail: Google Authenticator is TOTP or HOTP, depending on how the service owner set it up.

Can "same-machine 2FA" be still called 2FA? Where is the independent second factor if it all runs on the same machine?

Factor 1 is the password, factor 2 is an ephemeral 1-time-use token.

If the token's seed is secure (YMMV), it doesn't matter if it's on the same machine.

A practical attack against users is malware on the machine, which would allow screen viewing and key logging

An attacker can then compromise both factors at the same time.

To me, this weakens the point of 2FA a lot.

And another one is to try passwords from previous database dumps against your email/etc. This is even more effective then using malware.

If you are able to use a second device for your 2FA, great but it is not always possible and 2FA on the same device is still better then no 2FA at all. I don't want to carry around a key just for all my 2FA passwords but still want to use my phone for common tasks that require 2FA.

It's pretty irrelevant nowadays anyway. People log on their google, facebook, etc accounts from their phone as well as their desktop so for a lot of user, 2fa is done on the same machine anyway - that machine just happens to be the phone.

Additionally, the keylogging threat is severely reduced against most 2fa (because the login has to happen before the user sends the OTP) and can be eliminated completely by ensuring the token code is on a separate screen and creating a 30 second lock on the account while the user is entering the code.

Mobile device security (well for IOS anyway) is waaaaaaaay better than PC security (for ordinary users), so I'd disagree on that point that app+phone is the same as browser+PC.

In what world is malware which has full control of the PC stopped by having the token code on a separate monitor but the same logical system (if that's what you're talking about).

An attacker with control of the PC can just get access to any app. on that PC...

Maybe on a $300 phone.

This is fun, I have a tablet here on my desk I was given to repair a couple days ago. It's a mainstream tablet here in greece, sold for 70-90€, the sort of thing you'd buy if you are an "ordinary user" who doesn't know any better.

Well it's running a 4.x android with a custom marketplace and it is FULL OF ADWARE. Seriously. There's fullscreen porn popup ads on the OS itself. There's hundreds of spam notifications, a "fake" screenlocker which captures your taps into ads, etc, etc. Seriously, this thing is like a Windows XP SP0 PC from 2005 after heavy adblockless porn usage with flash and java on. I actually haven't seen anything like that this decade.

This is what you get if you buy an off-the-shelf tablet or phone here. This is what "ordinary users" get. We gave up control of our own devices in the name of security, yet shit like this can still happen.

(Sorry for the OT rant. Seeing that tablet pissed me off.)

Your confusing Out of Band authentication with 2FA. They are not mutually exclusive, and solve for different issues.

How does that mitigate against the one machine being stolen?

The convention is that the chances of two devices getting stolen together are smaller. That's why offsite backups are a thing.

It can, eg. if the seed is encrypted, but that's not what you're warding against. If the machine is stolen, one might install a custom CA or even custom web browser to MITM any https connection and just steal all your credentials, sessions and what not.

When Battle.net authenticators were introduced, the primary threat was keyloggers and leaked passwords (from reuse). Both of those threats are eliminated by TOTP, wherever it is.

If the machine is stolen and someone changes your software (browser, CA, whatever), that only hurts you if the machine is later returned to you and you don't wipe it.

But this discussion was more about if the machine was stolen and the bad guys logged in to your online accounts from it.

In that case it is true they would have your OTP, but they shouldn't have your Blizzard password.

It doesn't.

Passwords (on the OS account and on the Blizzard account) mitigate against the machine being stolen.

OTP mitigates against the password being stolen.

It is what you know and what you have so that if someone gets one they don't automatically get the other.

As an aside: The "What you know and what you have" paradigm is failing, imo. We often say passwords should be autogenerated, unguessable and unlearnable, stored in password safes like Keepassx. This removes "what you know" and you end up with "two things you have" (An OTP seed and a password in a safe).

When both the password and the TOTP seed are on the same machine - assuming you autogenerated your password - they are both autogenerated and out of your sight so they really are just two things you have, and you have them in the same place. The difference is, it's impossible to capture the TOTP seed from the TOTP token being input - passwords are vulnerable to that.

Am I crazy or does that make the password completely redundant?

If you lock your password database with a long, secure password that you know, then you are still doing it "right".

If your password database is not encrypted with a strong passphrase, then you are right, they both become "what you have".

Ah, that's true.

NIST does make a difference between the two, which is why it calls it "out of band authenticator."

I'm confused. Both of these are just "something the user knows," right? Making this explicitly not 2-factor.

Or, are we treating the ephemeral 1-time-use token as something the user has?

edited: But if your machine is compromised, attacker gains access to both factors.

edit: I don't know. Maybe your implementation meets the definition of 2FA. Regardless, it certainly reduces the security

Your problem isn't with 2FA, TOTP or SMS, it's with sharing an account that isn't meant to be shared. When you use a feature in a way it isn't meant to be used, sometimes you're going to get some sub-par user experiences.

Should Blizzard support multiple users of the same account? Perhaps, but that's out of scope here.

The reasoning behind why I wrote the library is completely and utterly irrelevant. I regret mentioning it.

Reply on news.ycombinator.com