Why Intel’s “How Strong is Your Password?” site can’t be trusted | Hacker News Reader

Why Intel’s “How Strong is Your Password?” site can’t be trusted

arstechnica.com

37 pointscoloneltcb42 comments

The worst part of this is that a lot of their advice for stronger passwords is idiotic and dangerous. They've taken methods that do increase entropy, but castrated them by making them systematic and predictable.

1. Using a multi-word passphrase is smart, but only if you generate it randomly. A novice reading their suggestion might think they can come up with a phrase that's meaningful to them. This is very bad advice and will lead to weak pass phrases, guaranteed.

2. The choice to delimit the words in your pass phrase with spaces/hyphens/title-case adds less than 2 bits of entropy to your password. If you do it the same way that everyone else does it (because Intel told you to), it adds zero bits. Randomly mixing in capitalization increases entropy. Predictably mixing in capitalization does not.

3. Adding numbers to your password is a red herring. The most likely effect that advice will have on a user is to encourage them to choose a pass phrase where a number fits in naturally, which will drastically reduce entropy since it constrains the types of phrases they might choose from.

And again, randomly throwing in numbers increases entropy. Predictably throwing them in does not.

4. Adding an exclamation, period, or question mark to then end of your pass phrase adds, again, less than 2 bits of entropy. Once again, random punctuation increases entropy, predictable punctuation does not.

And the icing on the cake is that their example "My 1st Password!" reinforces every possible bad interpretation a novice could make of their advice. Intel should really be ashamed of this piece of work.

What, so OnionMan77 not so good then?

http://www.theonion.com/articles/onion-twitter-password-chan...

I can't find it now, but my favorite such site was one that just took your password and replied with, "This password is completely insecure, because you just gave it away to a random web site on the internet."

Also, this has been making the rounds: http://www.ismytwitterpasswordsecure.com/

That was the first thing that came to mind as soon as I saw Intel's thing on security.nl. Then someone commented it even sent a request to the server with every check and I nearly fell out of my chair

There's a number, here's one of the better ones: http://www.inutile.ens.fr/estatis/password-security-checker/

I always remember this site for their Falso prover suite: http://www.inutile.ens.fr/estatis/falso/

That's the one I was thinking of. Thanks for the links. That Falso system is fantastic.

Yeah, I was wondering at the time why someone didn't set up a site like that for "checking passwords" but really uses it to build up a database of good guesses for offline cracking when you've nabbed the hashes.

My password is eyh7E6y1unZdyA3489dE

What are you going to do with it? That's right: nothing.

This webpage does not send the password home (confirmed with Wireshark). Even if you are under a MitM attack, this site would be the LEAST of your worries. This article is mere sensationalism and should probably be renamed "Why HTTP can't be trusted." What's that you say? HTTP open to MitM? Never!

I liked the fact that it complained about HTTP more than the fact that it's encouraging people to type their password into a random online site just 'cause it's by a "legitimate entity", whatever that means.

(Recommendations not to use your real password notwithstanding)

Really? If I were MITMing a person who accesses this site, I would make sure to tamper with the page so that it does send the password to a server under my control.

CONGRATULATIONS! It would take about 82432708400 years to crack your password.

While I agree with the sentiment in the Ars article, it seems to suggest that your password is submitted to a remote server. The check is completely client-side though.

http://www.intel.com/content/dam/www/public/us/en/apps/passw...

That's not their point. An unsecured site can easily be spoofed; a secured one would require either getting intel's ssl cert or convincing users to click through a big "dont trust this site" page in their browser.

The fact that they don't send the password just means that a MITM needs to put in slightly more work.

a secured one would require either getting intel's ssl cert or convincing users to click through a big "dont trust this site" page

Or breaking into their DNS registrar and getting your own cert for their domain, or breaking into any one of the many many dozens of CAs that most browsers trust...

Fair enough. Those are still generally on the order of very difficult, and certainly more difficult than doing none of those things at all.

It really liked "correct horse battery staple", but it did not like "hunter2" at all.

Yeah, I'm pretty sure those are both in every password dictionary ever...

This site would be much better if it took the time to "crack" the password you submitted.

Thousands of dollars of computing resources thrown at users of a free service would definitely make it better.

But you're giving them a plain-text password. It doesn't need to actually crack it.

This[0] was mentioned elsewhere in the thread. It's a good approximation of what I meant.

I realise putting "crack" in air-quotes probably didn't quite convey my full meaning.

[0]: http://dl.dropboxusercontent.com/u/209/zxcvbn/test/index.htm...

The thing about "correct horse battery staple" is that it has 28 characters including spaces. Sure, all of the words are common, but good luck trying to brute force it.

It's not possible to look at a password and determine how long it would take to crack it, without knowing how it was generated. All you can say is how long it would take _you_ to crack it. But just because a password looks hard to YOU, doesn't mean it's not a dictionary word in Bulgarian or something. There's no way for you to know.

It's also not possible to look at a password and determine long it would take to crack, even knowing how it was generated (to the reasonable extent I believe you implied, ie Bulgarian or English). Your password may happen to be the first word in a dictionary list or it could be the last word.

It is an inherently inaccurate probabilistic estimation of an attacker's methods, and does make any claim of being more than that.

If you know a password generation algorithm, you can compute a lower bound on the average time to crack a password given that it was generated by that algorithm. Some individual password might happen to get unlucky, but you can still have a good general idea of the strength of a password if you know the set it was drawn from and the distribution.

Apparently the password "significance" would take 317 years to crack, but "interpolated" would only take about 6 hours.

Both are randomly selected 12-letter dictionary words (from /usr/share/dict/words on Ubuntu, excluding words with uppercase letters or punctuation).

It's always funny to see things from HN comments come back to HN in the form of articles a short while later. Is it just that the HN crowd are on the ball or are the journos reading HN comments?

I'd say at least most HN'ers are on the ball. I glanced over that original headline and thought, "gee, let's go test our password by giving it to someone we don't know, that sounds like a good idea" (sarcasm detector explodes).

They do say "PLEASE DO NOT ENTER YOUR REAL PASSWORD", though admittedly it's in faint text.

I knew it couldn't be trusted, so I gave a spurious one (xkcd style), and it said that it'd take a majorly long time to guess. (But entropy per character was low, it was all lowercase!)

Regarding password strength checkers, see also https://github.com/lowe/zxcvbn

The following has a working example of that repo. I really like the implementation, and it gives wildly different results than the Intel site. http://dl.dropboxusercontent.com/u/209/zxcvbn/test/index.htm...

And as always, if you're going to enter your password into a page running zxcvbn, make sure to either disable your internet until you've closed the page, or disable plugins and monitor network requests to make sure what you type in isn't being sent anywhere.

Or you know, better yet, just don't type your real password into it.

And it claims "asdasdasdasdasdasdasdasdasd1234" would only take 12 seconds to crack.....

That surprises me, since it says "AAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" would take about 10^28 years.

Do you disagree?

Yes, i don't think its possible to crack a password that long within that time frame, am i wrong?

Yeah, you're wrong. You'd be right if password cracking was limited to checking 'a' through 'z', then 'aa' through 'zz', then 'aaa' through 'zzz', but that's not how it usually works in practice.

In practice, a cracking program is generally going to have a dictionary and an algorithm for generating passwords based on modifying and/or combining the words in that dictionary according to certain rules. The dictionary isn't just English words, but also common non-English words, words formed by keys that are close on the keyboard, etc.

If that dictionary contains "asd" and "1234" (which is a pretty safe bet), it will probably end up trying "asdasdasdasdasdasdasdasdasd1234" much sooner than you would predict based on an exhaustive search. I can't say for sure that 13 seconds is the right answer, but I think it's the right order of magnitude. We're talking seconds/minutes, not years.

Edit: changed "sequential search" to "exhaustive search" to reflect the fact that it doesn't have to be strictly ordered.

Thanks for taking the time to explain my comment.

I would've gotten to it eventually, but you've done a better job than I would've anyway, so it all works out. :)

EDIT: And just so we're on the same page... This is for recovering the plain-text password from a known hash. If you have to test each candidate against a web-service or something it'll take a significantly longer time.

Yes, I was thinking more along the lines of hacking into something rather then decrypting a password hash.

I've always liked Steve Gibson's site for entropy: https://www.grc.com/haystack.htm

I used a long dictionary word and it took them 5 months to crack ...

éxÇÏen0ôwâBümèzSäWgÎ.ÄîoqëÕêiCßÒ

Come at me bros.

Reply on news.ycombinator.com