Another downside is that changing master password means you need to change passwords for all sites or have multiple master passwords.
Another is that an attacker can use the knowledge of your password on site A to bruteforce your master password offline.
And so on...
Re. portability, KeePass and its siblings are compatible with every major platform.
You can even sync your database to all of your devices by storing it inside an encrypted volume for extra security.
This might be a concern if you're the victim of a state-level targeted attack. That's not a threat most of us have to deal with. Hackers aren't likely to spend a lot of effort cracking John Doe's password list; they want to steal a few million password hashes and sift out a few thousand easy or re-used ones.
Your other points about the weaknesses of the scheme stand. Still, if we could get more users to use not-terrible, not-duplicated passwords, even with a flawed scheme like this, overall internet security would improve immeasurably.
[0] https://www.theverge.com/2012/8/6/3224597/mat-honan-hacked-a...
Did those attackers guess or compute even one password at all?
There are many types of targeted attacks on passwords that don't come close to including state-level actors. Divorces. Corporate Espionage. Any one of the people on Judge Judy who posted information online about their co-workers.
With this scheme ANY site where you register can attempt to brute-force your master password offline. I fail to see how it's a good scheme.
Let's say you're master password only uses letters, numbers, and special characters. Just counting keys on my keyboard, there are 94 such characters. You should pick a random sequence as master password (very important).
Let's say you use the Antminer S9 (which can compute 1 gigahash per joule). For ease of analysis, let's say you can recognize the master password instantly. Also, say we're paying $0.2 per kwh. Then we can define the average cost c of finding the master password as a function of the master password length l: c(l) = 94^l/(1.8*10^10)
c(5) is about 40 cents, c(6) about 40 dollars c(8) is more than 300k, c(12) = 26e12
In comparison, the estimated amount of money in the world (in 2009) is 52e9 dollars. By the way, this is if you use a single SHA256 hash. You can make the hash arbitrarily expensive by iterating (computing h(h(h(master_pass)))).
The one and only argument against using a master password that is used to derive passwords is the single point of failure. If someone catches you typing your master password on video, you're pretty much fucked. But I guess this is the same for password managers.
There is still some truth in your arguments. I'm curious what you consider a good alternative?
How does one even calculate H(username|service name|master password) on an unfamiliar machine that might not have a decent scripting language installed on it? If you asked me to do it on my father's laptop, for example, I'm gonna have to download Python first.
Just spitballing which systems I would realistically want access to passwords on, at a minimum, includes: personal desktop, personal laptop, tablet(s), cellphone(s), family (parent, sibling, in-laws, etc) computers, office desktop, office laptop.
In my case (many cases?), the latter two prevent software installation, so I would need to manually type from a manager synced on my cell. Which really is no different to what is required for 2FA - just a longer character string. Overall, a some setup and synch related inconveniences but not to a damaging degree, which is why I think this is more a matter of perception - once you think through where you're typing passwords it appears less of an obstacle.
Cellphone and tablet I install the relevant app once and am done.
Either way, web access to the credentials isn't uncommon, though you'll need a device for 2-factor confirmation it's you.
For family devices...I wouldn't want to log into them. I don't trust their security sense and don't want to risk compromising my stuff because of their mistake. Why would I anyway; I have my cellphone on me, at minimum; I can access the account there.
I don't know about other Password Managers, but LastPass lets you access your vault using any web browser. So long as your network doesn't block the LP site, you can access your password manager from anywhere.
If I need immediate access to any other site and I don't have access to my password manager, I'll just trigger a password reset and check my email.
If it is a password I need to type, I use diceware "Correct Battery Horse Staple" style passwords. If I don't remember the password, I always have my phone, where I can retrieve the password from 1Password. There is no relationship between my different passwords. As long as my password vault is secure, so are all my passwords.
With your system, if I learn your master password and hashing mechanism, I basically have all your passwords.
On the other hand, if you take any other approach, you place either confidence in the security of a service (1password), or you store passwords somewhere. In both cases, you have a problem when you can not retrieve the passwords anymore.
If 1password is hacked or simply offline, or your phone gets stolen, you basically lost access to all services.
Security through obscurity. You're essentially just using the same password on every site now.
If you use a sufficiently good hash function with a salt to stop rainbow table based attacks and key strengthening for good measure, then an attacker who has one compromised password of yours isn't going to be able to reverse the hash to find what your source password is so that they can generate new per-site passwords using it. That is assuming they even try: how would you know that an arbitrary string of characters was generated that way rather than just being randomly picked by a password manager?
Of course all your accounts are compromised if someone knows your master password, but that is little different to an attacker getting your pass-word/-phrase for an online password manager that doesn't use some form of 2FA.
This is not a cryptographic attack on the hash but on the text that is being hashed.
These are two different classes of attacks, both being real-world attacks. The attack being discussed is against a suggested formula to generate passwords.
From the top of the thread, "a scheme that hashes your username, service name, and master password to generate a password." Discovering the password by determining the inputs to the hash function is all that is needed. This is the case even if a different type of function is used instead of a hash.
Apologies, if I gave a different impression.
Introducing a good salt would stop that but then the salt would need to be stored and managed somewhere (securely synced everywhere you might need to regenerate a password - it can't be a public value or you are back to square one with the dictionary attack). You could also introduce a CPU and memory intensive key derivation function to frustrate the brute force attack a little but that is a race you won't win if you need the function to work on a low resource device and the attacker has a large cluster of processors and/or plenty of time to work with.
It would protect against password reuse attacks even without the salt: if site1 stores plain passwords and leaks them then getting into my account on site2 becomes no easier than it already was if I had different passwords for each site (we are still looking at a brute force dictionary attack against the master password) so the scheme could work if the user has a sufficiently complex master key (and never needs to change that key). Making the master key complex enough but memorable is the challenge there, back to one of the core password related problems but at least with only one (or a few) to remember not one per account. Changing the master key once you believe it has been compromised would be quite some hassle in a similar way to if your master password for a service like 1password were compromised (worse: with a service like 1password you at least have a list of accounts to work down but with this scheme you need to remember them).
An interesting though exercise, but I'm probably overthinking an unrecoverably weak idea...
This is a very bad mechanism to create passwords that will create severe security vulnerabilities. You are adding complexity and false-security without benefit.
Passwords are a secrecy mechanism, not a formula mechanism. In your method, all one needs is the formula, then all passwords will be disclosed.
The suggestion might then be to interject a secret with enough entropy into the text being hashed. But, that is no different from a secret password itself!
A better way is to use a scheme that hashes your username, service name, and master password to generate a password. A problem is that this doesn't always comply with the arbitrary demands on your password.
This is why these arbitrary demands need to die: they make the only way to securely and conveniently access accounts from different devices impossible.