[flavor8]

> Breaking the code, according to an Apple technical guide, could take “more than 5 1/2 years to try all combinations of a six-character alphanumeric passcode with lowercase letters and numbers.” (Computer security experts question that figure, because Apple does not fully realize how quickly the N.S.A. supercomputers can crack codes.)

Uh, what? Surely the journalist has missed an important technical detail here, right?

https://www.grc.com/haystack.htm

[sopooneo]

I had the same thought. If we allow upper, lower, and digits, that's 26+26+10 = 62 possible characters per space. With six spaces, that gives 62^6 = 56,800,235,584 possible passwords.

Now if we take their figure of 5.5 years to crack a phone's files and divide, we get 327 seconds (more than 5 minutes) per password they check.

Something is off, though perhaps it's my math so please do double check it for me.

Edit: Argggg. Good corrections. My main problem is that I did my final division in the wrong direction. Fix that by taking a reciprocal: 1/327 = 0.003 seconds. And then correct that by a factor of 2 to assume they get each password in half possible time: 0.003 * 2 = 0.006 or roughly 6 milliseconds. Thanks for the quick check folks.

[lozf]

6 randomly selected characters out of the 62 available gives less than 36 bits of entropy, which anyone with even a passing interest in any kind of cryptography will instantly recognise as pretty poor.

This issue is compounded by the fact that humans are notoriously bad at randomness. I really don't think many users will be typing the 22 random characters required for just over 128 bits of entropy every time they want to use their phone.

But maybe the 5.5 year figure includes the incrementally increasing delay that Apple insert between tries after x wrong guesses -- assuming a manual brute force, which is pretty much not how it would play out in reality.

[XorNot]

In reality they lift your prints from the phone, fool the lock sensor, then clone it to a new, bugged phone, and monitor all your communications.

But frankly, such scenarios are not privacy concerns unless you're actually trying to carry out crimes because at that point you've got a half-dozen or more government agents assigned to personally follow you.

[Ntrails]

5.5 years is ~173563000 seconds or my maths and google are both wrong.

For pure brute force you'd want to make some assumption about the mean time taken to find the correct password, but lets argue that you find it after checking exactly 1/2 of the possible combinations.

2.8400e+10 passwords checked in 1.7300e+8 seconds => ~0.6 milliseconds per check. ish. I think.

Edit: as comment below points out, humans are crap. In reality any hack would use dictionary attacks rather than pure brute force. I was just addressing the maths.

[theGimp]

There could be truth in that number depending on what encryption algorithms are used. Some take significantly longer than others to apply.

On a tangent, I trust Apple's competence but I have seen no information on the technicalities of this feature. I would like to hear specifics so we can estimate how secure the design is based on facts.

[ghshephard]

Re: "I have seen no information on the technicalities of this feature."

https://www.apple.com/privacy/docs/iOS_Security_Guide_Sept_2...

[privong]

> There could be truth in that number depending on what encryption algorithms are used. Some take significantly longer than others to apply.

The computational burden likely cannot be too high, as it is running on mobile hardware and people expect rapid access to their data after entering their passcode.

[rdwallis]

The decryption is probably being done by specialized circuitry on the chip. It might perform the operation faster than a much more powerful machine.

[privong]

Good point, thank you.

Going back to @theGimp's comment, it is probably safe to assume that nation-states that might want to decrypt this would have some form of access to that circuitry (either a specialized implementation or by using the apple chip) and could sidestep the speed issue. It seems like, to 0th order, it would either be slow for everyone or quick for everyone.