Hacker News Reader

Comment by verbify | Hacker News Reader

LastPass always represented a single point of failure for me, although still bettter than what people usually do - same password across all services.

I recall one interesting method involving no third-party services is a concatenation of a random string combined with some of the letters of the service I'm using. E.g. the first and last letters of the website I'm on. So if the random string is xPv4rz and I'm on HackerNews, the string becomes hsxPv4rz.

View on HN

The problem is that you don't have control over password requirements, so you'll rapidly accumulate exceptions. What if they require (or assign) a PIN? Security questions? What if they assign a number as your login? Require a special character? Forbid special characters? Require >6 chars? Forbid >6 chars? Forbid your password from matching previous passwords? What if they don't show you the domain? What if the domain changes during login (see: Cambridge Savings Bank)? What if the domain changes over time? What if the company name changes? What if they lock you out after 3 attempts and you burn up all three trying to remember the variation you were forced to use?

I tried the combination scheme for a bit over a year. I was in denial about how poorly it was working until I started getting politely teased by casual acquaintances about being the guy who could never remember his passwords. There are accounts I'm still permanently locked out of. It was bad.

By comparison, LastPass has been great. More secure passwords, less forgetting them. Eventually I want to move to a password archive that's physically separate from the computer, but for now I'm happy with no longer dreading login screens.

Well regardless of your scheme, you should still write down your passwords!

Just because it's not feasible to lug a moleskine with that long long list of passwords to everywhere as a "password manager" solution, doesn't mean that you shouldn't write them in that notebook, which you store in a safe place. Not for quick & easy access, but for backups.

You don't have to do it on paper, but it's a good choice because it's low-tech, easy for anyone to use (bus-factor, next-of-kin) and exactly as secure as your common sense, making it exactly as secure as anything else you can use.

Being unable to access accounts, because you forgot the passwords, that you didn't write down anywhere is equivalent to losing data because you didn't make backups. Which we also tease people about, right?

Have you written down your LastPass master-password anywhere?

The tone of your lecture suggests that you think I disagree. Which is odd, because from where I'm sitting your central claim looks like a corollary to my conclusion.

The main issue with this kind of method (using an algorithm to create a password) is that it makes it very hard to change your password if needed.

E.g, let 's say you use your algorithm for your passwords : Amazon get AnxPv4rz, HN get hsxPv4rz, etc... Now, someone get your password for Amazon one way or another (social engineering, poorly protected website, etc...). You have two choices :

* Change your algorithm for all your passwords. Depending on the number of different websites you use, it will probably be a very heavy process, and you might forget some websites.

* Use a different algorithm for Amazon. If you start going this way, you'll quickly find yourself with 5-6 different algorithm, each more or less complicated. That's pretty much why people use Password managers : they want secure passwords, without having to remember each and every one of them because it gets complicated after the 5th. And if you use a password manager, you might as well use completely random passwords.

Additionally, if a hacker gets 2 of your passwords, they can see that all you passwords are of the form AxxxxxB, from there it is quite easy to get all your passwords. If you have an "algorithm" that is easy enough to remember, then it is most likely breakable by a smart hacker.

It would not surprise me if a single password is enough for an attacker to guess the rest in this scheme. It is common enough to be predictable, and even if it weren't, all your passwords are at most 2 characters different from each other. That's not a lot of entropy.

A possible mitigation of the obvious problem (the pattern itself) is instead to use a hash function on some aspect of what you're trying to log in to: a hash of the main domain, the human name, whatever. That way if a password is compromised, it may look like a common pattern is being used, but the "random" part of one site's password won't do any good on another site.

The challenge being to implement a hash app everywhere you might be that will generate the correct hash. Maybe a very simple mental hash function would be enough; mass attacks would probably be foiled, and individual attackers may be too lazy to bother figuring out the hash.

There remains the problem of changing the password due to some necessity, like policy or compromise.

EDIT: If this was app-based, then each time you change a password, the app would record that fact, and incorporate it into the hashing data, so that a different hash is generated. The problem then being synchronizing that fact among all your installations of the app. And if you were using a mental hash, I don't know how you'd remember this across all your logins; possibly some central piece of data that you could easily look up, to see that you've changed your password one or two or N times, that is easy to incorporate into the mental hash.

This is good brainstorming, but I'm getting tired just thinking about all that mental complexity. I would rather take my chances with Lastpass than keep a complex mental list of how many places I've reset my password at so far.

I've actually been using this approach for the past 12 years. There used to be an old firefox plugin called Password Composer (http://web.archive.org/web/*/http://jlpoutre.home.xs4all.nl/...) which would let you type in your master password into firefox, append the website domain, take a digest of the result and truncate to 8 characters before sending to the website. Over the years, I replaced this plugin with a simple 5-line script so I can reproduce my passwords from anywhere, strengthened the hash function from SHA-1 to bcrypt, and also added a configurable password length so I can gradually grow passwords over time as computers grow more powerful. Here's my solution: http://akkartik.name/pc

Regarding other comments in this thread about managing exceptions to the scheme over time, I have a simple solution: I let my browser save my passwords. I think this is a reasonable trade-off; the dominant threat online today is not somebody gaining physical access to your devices, but random script kiddies brute-forcing passwords using stolen hashes.

Tl;dr - use my script to generate passwords, use your browser to remember passwords, and the complexity is quite manageable.

Edit - sorry, I forgot that that 5-line script calls another 5-line script: http://akkartik.name/bcrypt_digest.

One potential gotcha I realized recently is that it uses the bcrypt encryption scheme rather than the well-known digesting algorithm. The two share a common kernel, from what I can tell, though there are significant differences. There might by cryptographic implications to using an encryption scheme for digesting; I'm not an expert. Hopefully the fact that this is a bespoke arrangement being used by a tiny population will keep anyone from hacking me :)

If you're storing your passwords in the browser anyway, why not just generate random, arbitrary passwords anyway?

It gives me a way to figure out my password on a strange system. Admittedly this use case is growing less common with the ubiquity of personal devices and sync.

So as your browser stored passwords pass through (and rest at) the server, waiting to be synced to another browser or device, you have passwords "up there." Are they in the clear? Are the encrypted well enough?

https://blog.mozilla.org/services/2014/04/30/firefox-syncs-n...

Your point is well-taken that I'm just replacing one external service with another, so the difference isn't large enough to convince anyone to switch. My original response in this thread was just to point out prior work relevant to the brainstorming above.

> I recall one interesting method involving no third-party services is a concatenation of a random string combined with some of the letters of the service I'm using. E.g. the first and last letters of the website I'm on. So if the random string is xPv4rz and I'm on HackerNews, the string becomes hsxPv4rz.

I don't see this as really that useful -- if someone gets your password from two leaks they can probably figure the pattern out as well as you can.

That seems like a lot of work given the number of passwords attackers get when they compromise a site. But then again there are scenarios where it would be a concern. I actually do use this technique but only for my least secure tier of logins.

Isn't using the same basic string with only a few tweaks to it here and there easy to crack (especially if the delta is information related to the site it's for)? And doesn't it represent its own sort of "single point of failure"?

I'd say as long as you only have to remember one terrible string of junk, it might as well be xPv4rz$23aAjQ1&oD rather than xPv4rz.

Reply on news.ycombinator.com