Without backups, having a phone die or get lost is a very frustrating experience. If you use backup codes, you have to go re-register 2FA on every account with backup codes. And of course some don't offer backup codes.
You could screenshot and print out the QR code at registration time (I've done this for a few accounts), but shared printers apparently retain history of everything they've printed, so you'd need to own a personal printer, which seems absurd.
You could write down the TOTP secret on paper instead. Ideally with multiple copies and then move them to different physical locations. That's a big hassle to do for each new account registration, which seems to happen every few months.
Encrypted backups solve this easily: you get one key, which you write down once and distribute to different locations if you want. After that there's nothing new to do for new accounts at registration time. And restoring onto a new phone after one dies is also easy.
However I do have an issue with 1password's feature of auto-filling those codes, seems like it's just invalidated the whole "something you have" party of MFA.
For me Authy is a happy medium
It isn't that bad, it still a 2nd factor, but it's worth considering it.
I built exactly this: https://github.com/yeo/bima
I stored everything into a SQLite in `~/.bima/bima.db`
Your OTP secret is encrypted using a master password that you chooese. I use AES GCM for encryption: https://github.com/yeo/bima/blob/master/shield/encrypt.go#L1...
You then has 2 options to backup/sync among device it:
1. backup that file using dropbox, icloud, google drive 2. Enable sync to my backend. You sync encrypted data, even me cannot see it. Then other device can sync from my backend, and you enter your master password to decrypt it.
Entire thing is open source, implement using Golang Fyne UI toolkit so it run across linux/mac/window/ios/android.
If you want to help beta test it, I can send you a beta build.
Also, part of my ideal requirements is an app built by an entity that I trust as much as Google. Open source is great, and this app is simple enough that I can skim the code once, but I'm not going to do it for every update, and I might miss something. There's still something to be said for that kind of accumulated trust. (I might not trust Google as much as I used to, but I still keep my life on gmail, so I have to trust them pretty far.)
Biometric locks are interfaces that the OS does not expose to users, and that are backed via an HSM. On the iPhone and modern Android, they allow you to envelope encrypt a message via biometrics in such a way that it can only be unlocked from within the app that locked it AND via enrolled biometric signatures. Passcodes are the same, but easier to phish.
It is incorrect to say that they "are a gimmick and offer negligible value".
In this document, they lay out these mechanisms in great detail. Anyone with an interest in security should read this, imo. It's a magnificent "blue team" effort https://www.apple.com/ca/business-docs/iOS_Security_Guide.pd...
I'm well aware of how the mechanism works.
If the host OS or the app is compromised, it won't help - the attacker can just steal the codes after they're unlocked by the enclave.
For physical theft, it only provides a meaningful advantage if the phone is unlocked. Someone stealing your unlocked phone to get at your 2FA codes is well outside of most people's threat model.
If the host OS or app is compromised, then it absolutely helps! The codes cannot be stolen until the user unlocks.
Considering that the most recent iOS jailbreak was not persistent, due to their chain or trust, this is absolutely helpful. It's one component of defense in depth.
In a system with a properly established chain of trust (as ios and android have now), compromising only one component is not enough to ensure complete user compromise.
In your "advanced threat" example, where a fed compromises a device, powering off and then on the device will hopefully put the device into a trusted state again. Aside from the boot0 exploit that nailed ios a few months ago, a few revealed state-level attacks _were not persistent_. Even here, it helps!
>For physical theft, it only provides a meaningful advantage if the phone is unlocked.
These biometric mechanisms are implemented in such a way that "opportunistic" theft (such a as a mugging) yields a positive outcome for average users.
>Someone stealing your unlocked phone to get at your 2FA codes is well outside of most people's threat model.
Aside from the aforementioned compromised host example, which has been a problem on android in the past.... 2FA, sure, that's a bit more esoteric... except... not really. Since 2FA can be used for payments, or other sensitive info, we should probably give users the opportunity to protect against. I work for a personal finance company and we enforce biometrics for a 2nd factor on login to help protect against both the compromised host threat, but also to prevent against spousal, familial id theft (more common than you'd think, sadly). Your familial member may know your passcode, but they will be unable to access your financial data.
I am someone who would be considered to be a "high value target" since I have had some of these things happen to me and I'm certain they will happen again in the future. I work in a politically sensitive industry where it turns out these things are commonplace.
When I was younger I was always security conscious and took serious trade-offs to maintain my privacy. You could have called me paranoid, but I think I was more excited at the idea of being protected against all of these high-level threats, even if most of them were nonexistent at the time.
While this discipline certainly benefited me later in life, I've realized that I will never again get the chance to reasonably evaluate my threats as "average" and enjoy some of the simpler conveniences of technology that come at the cost of compromising some high-severity threat models. Of course I still think there are a lot of precautions the average person can take to protect their privacy without major trade-offs, but those don't usually include state actor-level threats against your devices.
I wouldn't change a thing now, but when I talk to people who see no reason to take anything more than the average precautions, I no longer think it's "cool" to do anything more than that because I know for a fact I would too if I was in their shoes.
I think @trickstra was quite civil in their comment. And they also brought up a completely valid point:
> > And you cannot change it like a password. So that's why it's a gimmick.
Regardless of threat model, not being able to change biometrics makes them very high value to an opposing force. Using biometrics "for the masses" to whom they don't have that completely different threat model practically eliminates their ability to upgrade their threat model since their biometrics have potentially been compromised already. Therefore, it very much is a gimmick.
If your adversary is a random thief, or the untrustworthy general public, then it works great, and is a significant upgrade from the zero security that most people had prior to the proliferation of biometrics on phones.
Imagine a leak from a database storing biometric keys, is this a disaster on par with a normal leak? In my opinion it's even worse!
You might say we only use biometrics for a local authentication, but that would limit their application. If you convince people biometrics are great, such database will inevitably be created - what are the chances we store and guard biometrics better than we do with passwords?
This, by the way, still also requires physical access to the device if used for local authentication. And in the event biometrics are used in a physical location (which I’m not personally a fan of, but let’s consider it anyway), there’s often also a human there as well. Your clever mask may fool the sensor, but it probably also still looks like a mask to the guard behind the desk.
Really though, I think "one-time passcodes" is the wrong escape hatch because it is actually just password auth, not a second factor. The way it ought to be is that you hook up multiple devices. (e.g. like having both a phone Authenticator app and a USB key.)
Frankly, most of MFA is security masturbation, is nowhere near to being a real solution, not even a real problem. Its just neat from a technical standpoint (if done well, which nearly never happens).
A minimum of 2 registered u2f keyfobs, that's all you need to make it usable and safe. Nobody does it, and trusts people either backup their otp codes or never lose or break phones. Ridiculous.
I also went through writing a simple javascript page to import all those keys to Google Authenticator via totp:// protocol. It takes a csv of all strings, and spits out the html links with totp://protocol. All using local storage only. So whenever I change phone, I paste the csv from keypass to google keep (pretty safe), open it on the new phone, copy it & paste into my html5 app, and click each link and it auto adds itself to Google Authenticator. Then I select all text in Google Keep and delete it.
Essentially, what's your backup plan so if the very worst happens when you're in the middle of nowhere you're not completely locked out?
e.g. "What if I'm traveling, I lose my phone, and the country I'm in blocks internet access to my 2FA service?" "What if I'm traveling, I lose my phone and my wallet?" (In this case, even if you use a syncable 2FA service - how will you pay for a new device to access it without your wallet? You have a different problem here, that likely involves going to an embassy and/or relying on friends & family!)
The question is, (a) how _likely_ are those scenarios? And (b) how much risk/damage/pain do they cause?
If I'm traveling, I'm fine unless I lose my phone. If I lose my phone, then I guess I can't sign in to some/many things until I get back. If I lose my phone AND my wallet and I'm not at home, I'm probably more worried about getting my ID cards and bank cards replaced than accessing certain websites. If my house burns down with my phone, wallet, and backup cards inside, I'm probably just thankful to be alive.
My 15 year old Brother doesn't have that capability... So on that front, a printout of QR codes is more likely to be compromised by the paper itself being left out somewhere.
If you are worried about that then using a desktop computer in the first place is a bad idea.
Worst case, i doubt it would be that difficult to hand draw the qr codes on graph paper. Would be quite tedious, but they should be error correcting to some extent.
What's the threat model here? If the CIA is in my home network, I'm sure they can get into my email account too.
I find that a terrible design decision, because now every website has to implement their own backup code flow. PayPal, for example, is one of the most critical applications when talking about authentication, and while they support 2FA they don't give you any backup codes at all: https://www.paypal-community.com/t5/My-Account/Backup-codes/...
Now I cannot synchronize devices, I'm hosed if I ever lose my phone, and I'm still relying on PayPal customer support to do a proper out-of-band authentication. The worst of both worlds.
However, I was able to set up Authy as a secondary authenticator.
I don't have SMS as 2FA for PayPal.
Not sure if HN allows to plug your own apps, so please forgive: I made an app a while ago that aims to replace Google Authenticator for some of the reasons mentioned: it allows to back-up and transfer tokens without creating a large attack factor. Not having sync is a feature in this case as well. In fact, the app does not even have the internet permission enabled, so it utterly unable to phone home. Transferring backups does require a biometric lock.
It is also entirely free, so I'm only posting this out of pride of my own work: https://play.google.com/store/apps/details?id=com.pixplicity...
1pass user, while I know that it's not as secure as keeping them separated I view it as if someone has access to my 1password vault unencrypted I'm already screwed, but having 2FA coming from 1 tool still protects me from the case of my username/password pair getting leaked or MITMed.
So if someone got a hold of my LastPass master password, they can't access my google account, banking, etc.
The biggest security issue is that they hold the keys to my entire kingdom. So if someone somehow gains access to that account, I need another layer of protection, no matter how small the chances are.
https://www.theregister.co.uk/2017/02/28/flaws_in_password_m...
Mostly I'm making an appeal to authority to tptacek [1].
https://duckduckgo.com/?q=tptacek+1password+site%3Anews.ycom...
My 1pass vault is decentralized, and also includes a robust password generator for making stuff on the fly -- something Dashlane lacks. I'm not sure I'm using either service fully, but one of them is a work freebie and the other is from back before 1pass went to a subscription model.
... the idea isn't that either are perfect, but since password sharing and phishing are a bigger attack vector than whether my phone falls into the wrong hands, using 2FA as anti-phishing insurance is still better than not.
And it's easy enough to synchronize multiple devices to the same qrcode when setting up 2FA so that you can generate codes from a backup device if one goes missing.
I also switched to OTPAuth, great app, provides encrypted backups and switching to a new phone was seamless. Would absolutely recommend over Google Authenticator.
Source? I've seen some vulnerabilities over time, but the "giant attack surface"?
- Passcode or biometric locks on an app are a gimmick and offer negligible value.
- The keys not being backed up or or synchronised across devices is not a bug, but a feature. You're supposed to keep offline backup keys. Any sort of synchronization feature adds a ton of attack surface.
- In particular, Authy, LastPass and 1password have a giant attack surface compared to a simple app like Google Authenticator. They also rely on centralized services and if you also keep your passwords in there, you eliminate the whole point of two factor authentication.
- One important risk with authentication apps is compromised updates, and Google Authenticator has a very low risk of this since it's backed by Google's strict security processes.
What you should actually do is to move to U2F/WebAuthn and pester any of your service providers that do not offer it. Yes, if you need to replace a token, you'll have to go through all accounts and change it. There's is absolutely no way to prevent this without compromising on security.