Hacker News Reader

Comment by devx | Hacker News Reader

Isn't it about time we get new security standards for Wi-Fi? Is there anything in the works right now to replace WPA2?

There's nothing seriously wrong with WPA2 itself. I'd consider it as secure as pretty much anything else out there that uses 128bit AES (given that your key exchange is secure of course - read on below).

The problem is with the PSK variety, mainly that it's susceptible to offline dictionary attack: about 5% of actual WPA2-PSKs can be easily guessed [1].

There is stuff in the works to fix this though. My favorite is EAP-PWD [2]. It's resistant to offline dictionary attacks, it has perfect forward secrecy and it's already supported by Android. Basically, it's what WPA2-PSK should have been.

In the mean time, if you're security conscious just set a long random PSK or configure e.g. EAP-TLS. Both will give you strong security against pretty much any attacker.

1. http://wpa.darkircop.org/

2. http://tools.ietf.org/html/rfc5931

Is there a way to authenticate that you are connecting to your AP?

Yes, check out the many variations of the EAP and PEAP authentication protocols. Client authenticates the access point by its certificate, client gets authenticated by his/her client certificate. While some combinations have some flaws, like MSCHAPv2 may have too short keys for instance, there are others that I consider quite solid.

http://en.wikipedia.org/wiki/Extensible_Authentication_Proto...

EDIT: Sorry for having repeated some of the stuff bjornsing already said. Should read through more carefully before rushing to comment, I guess.

Yes, and even WPA-PSK (with a strong pass phrase) has trustworthy mutual authentication: your device will (or should) not connect to a rouge AP that doesn't know the PSK.

The problem isn't limited to WPA2. As far as I know, in and of itself it's actually fairly secure.

Most of the problem is that passwords are either easy for computers to crack or hard for humans to remember. The middle ground has disappeared as computational power has increased.

> passwords are either easy for computers to crack or hard for humans to remember

Obligatory xkcd comic: https://xkcd.com/936/

I loathe whenever people post that comic for one simple reason.

Although mathematically the password given in the comic has a higher entropy and would take more time to crack under normal circumstances, the problem is that it follows a very simple and easily describable pattern: smash (four) dictionary words together into a combination.

Crackers will simply start using wordlist rules to generate large lists of meshed together dictionary words and use them if they have good reason to believe you're using this pattern (pretty sure it's simple with tools like Crunch). Whether they'll guess the proper order is unknown, but as with any other case people will use certain permutations and combinations more than others.

The point is you're supposed to use truly random word combinations since those are at least memorable.

  $ wc -l /usr/share/dict/words
  119095
  $ python -c 'print(119095 ** 4)'
  201175048646341950625
  $ python -c 'print(85 ** 10)'
  19687440434072265625

So, even if your target is known to be using this scheme in pure form, this has more entropy than a completely random 10-digit password (assuming ~85 characters) -- and who would actually be using such a thing, except someone using a password management program - who could just as easily be using a 20-character random password?

So even if it becomes known, it's an improvement on what users are doing now.

Can never turn down an opportunity for a one-liner.

  $ perl -E 'open(my $fh, "<", "/usr/share/dict/words"); my @words = map {chomp; $_} <$fh>; close $fh; say join " ", map {$words[int rand @words]} 1..4'
  menu chemists administrative seeps

Might have to run it a couple of times before you get something that you can memorize.

You shouldn't use a non-cryptographically secure random number generator (perl's rand) in the context of password generation. It's too risky.

Ew.

    shuf -n 4 /usr/share/dict/words | tr -dc 'A-Za-z0-9'

You can use a dictionary of the most common 10000 words, you'd still have loads of entropy.

The whole point of the comic is that even with that known "simple pattern," you still get higher entropy than a normal password.

Obviously a random 64 character string would beat either of them, but if you're expecting a human to memorize a password, correct horse battery staple is clearly preferable.

I think you've missed the point.

There are more possible permutations of four words than permutations of 10 upper/lowercase letters, digits, and common symbols. The four random word approach is harder to crack.

I also add digits, some punctuation, a misspelling or two and the words are not in English. Oh, and I've got different passwords everywhere.

apriorixWasxTotallyxWrongxAboutxThis

is a really good password that people can remember easily.

Added to dictionary.

snap!

Fortunately, it is entirely unnecessary to remember your Wifi password (provided that you trust your devices…).

Create a near-random 63 char password, put it in a text file on a USB key and possibly print it out as a QR code and you’ll never have to worry about either entering it by hand or it getting cracked by that strange kid across the street.

until you buy an apple TV (and don't want to cable it). Fun ensues.

I use a random 63 character WPA2 password, and my solution was to cable it initially, and then set up the WiFi password using the iPhone remote app.

And if all else fails, entering 63 characters is not really that hard either (unless the Apple TV has one of these weird on-screen letter-choosing wheels you sometimes encounter in videogame consoles and the like).

Oh god, yes it is. Especially when you realize halfway through that the iPad cannot type "`".

(You can copy and paste it, though.)

Or Roku or a nest... Tons of devices use wifi but lack cameras or copy/paste.

The best human rememberable password is 4-5 words from a dictionary + a special character. Assuming that most people have roughly 20k words vocabulary and that most keyboards can type easily say 60 characters, you get 20,000^460 or 9.610^18 passwords. This means that if you were to crack at a rate of one billion (with a b) passwords per second (unrealistic) you would still take on average of 9,000 years or 18,000 years maximum.

If a person is choosing the words, they aren't going to be uniformly chosen from the list. For example, I'd be willing to bet "monkey" is much more common in passwords than "contacted" even though both words are probably known by about the same number of people.

Even so, of you assume that average person knows 20k words but would only use about 2,000 words in day to day life, and thus in his or her password. That still means that at one billion passwords per second (which is completely unrealistic, unless you are NSA) it would take 1.8 years to crack the password. Who would invest that much time and electricity into a wifi password, unless, again, you are the NSA :)

WPA2 is fine for what it's intended, provided you're using a long random key, otherwise the number of key-strengthening iterations could use some beefing up.

There are a few problems with all PSK schemes that make internal attacks problematic. Anyone who sniffs your initial handshake and knows the master PSK can read your traffic. There's a lack of mutual authentication. Having a scheme where each device registers its own password with the AP would probably be better.

Other than that, it's generally a good solution, why do you feel it needs replacing?

WPA2 is probably ok with a long passwor(d|phrase). The problem is nto WPA2 per se but another bundled technology in many rounters. WPS -- that is crack-able very easily. Many routers that even say they disable it don't really do it.

Search for it there is a list of routers that are better than others. With WPA+WPS we are mostly back to WEP days where any kid with a laptop and some googling skills can get access to many wireless networks.

Reply on news.ycombinator.com