Zini: Adjusting with Moore's Law

[Posted March 23, 2016 by n8willis]

Enrico Zini reports on his recent investigations comparing the strength of various password generators. In particular, though, he notes that the time estimates reported by strength-checkers like zxcvbn do not account for how computing power will likely increase in the hundreds of centuries they often report. Making that adjustment changes matters noticeably, he found:

If we assume that every 18 months the effective computing power will double [...] then something that is now estimated in 152 [centuries] is likely to take, well:

    $ ./actual_years 15200
    9.33 years needed

Zini: Adjusting with Moore's Law

Posted Mar 24, 2016 15:44 UTC (Thu) by spaetz (guest, #32870) [Link]

By coincidence(?) the recent Economics featured an interesting special on the post-Moore's-law time: http://www.economist.com/technology-quarterly/2016-03-12/...

Zini: Adjusting with Moore's Law

Posted Mar 30, 2016 6:51 UTC (Wed) by robbe (guest, #16131) [Link]

log₂(15200)·18/12 gives me about 20 years and 10 months.

The linked blog has already updated the estimate to 21 years.

Zini: Adjusting with Moore's Law

Posted Mar 31, 2016 12:23 UTC (Thu) by welinder (guest, #4699) [Link]

There's really a parameter missing, namely how often the equipment is replaced
with new and improved hardware. For simplicity, set that to once a year. That means
the new machinery is 2^(m/12) times as fast.

Here's a table showing how 15200 years becomes as a function of the number of
months used in Moore's law. Even at 36 months doubling time we still get there in
~35 years!

mth rate years
--------------------
12 2.00 12.9
13 1.90 13.9
14 1.81 14.9
15 1.74 15.8
16 1.68 16.8
17 1.63 17.7
18 1.59 18.7
19 1.55 19.6
20 1.52 20.6
21 1.49 21.5
22 1.46 22.4
23 1.44 23.3
24 1.41 24.2
25 1.39 25.1
26 1.38 26.1
27 1.36 26.9
28 1.35 27.8
29 1.33 28.7
30 1.32 29.6
31 1.31 30.5
32 1.30 31.4
33 1.29 32.2
34 1.28 33.1
35 1.27 34.0
36 1.26 34.8

In Gnumeric, that is A1=12; B1=2^(A1/12), C1=ln1p(15200*(B1-1))/ln(B1)-1 etc.