[Cryptography] "Anthropic A.I. Model Finds Flaws in Tough-to-Crack Encryption Algorithms"

[Peter Gutmann]

Nicko van Someren <nicko at nicko.org> writes:

>I think it starts at 10^4 possibilities (about 100 times smaller than 4^10)
>and if there are four abraded buttons it diminishes to P(4, 4) == 24, but
>your point still holds :-)

There's electronic versions of that too.  The majority of garage door openers
a few decades ago were built using the universal Princeton 2262 encoders or
one of its clones or equivalents.  The 2262 has 3^12 = 531,441 combinations
from encoding via tri-state address pins, only no-one every tri-stated them so
it dropped to 2^12 = 4096.  However some of the pins were multi-purpose and
also located on the other side of the DIP package so manufacturers of remotes
only wired up the 8 pins on one side for 2^8 = 256.  Except for the ones who
got cheaper 6-position DIP switches for 2^6 = 64, or who followed the
reference circuit with a 4-position switch and 16 combinations.

This is why your remote would sometimes open a neighbour's garage door.  For
extra effectiveness you'd get it wired into your alarm so it would disable the
alarm as well.

Also, the matching 2272 receiver had no anti-brute-force mechanism so you
could sweep the entire 256-value range in a few seconds.  I never understood
why these weren't used more by burglars, although possibly having every garage
door in the street suddenly open might be a bit of a giveaway.

Peter.