Back when these were declassified, I prepared a transcript and annotated parts of it: http://gwern.net/docs/1955-nash Good for those who don't want to go through the scanned PDF version but a HTML version. (My version also has links to various discussions such as class exercises in breaking his proposed system.)
There are comments here trying to contrast Nash's work to modern cryptography, as though that's somehow fair. Nash never had access to anything remotely like a personal computer and this is just after the age - and in fact still one - where physical and mechanical devices were being used to perform encryption.
The better thing to do would be to compare (rather than contrast) his work. Look at this letter. Nash essentially predicts trapdoor functions and the P/NP gap... in a mere letter.
We were lucky to have Nash and his contributions. It's unfortunate that most of his academic work is largely unknown and that pop science focuses disproportionately on his work on equilibrium in symmetric games.
I could be wrong, but I believe that Princeton PhD's go by "mister" instead of "doctor."
Princeton considers itself a unique institution in this way. They do not accept transfer students, they do not have any professional degree programs, etc.
The diagram on page 14 (page 3 of the machine description) vaguely resembles a Markov model, which I could imagine would be useful for cryptanalysis, but I don't see how such a system would be reasonable for encryption or decryption if you have the keys. Am I misunderstanding the machine?
The NSA's response said about the same thing, once you get to the end --- not interesting, not reasonable for encryption and decryption.
Interesting, I think, the statement that Nash's machine requires "comparatively" too much hardware. This implies the NSA did have auto-keying systems at the time that had higher security, and lower hardware burden.
Remember, in '55 we didn't have single-chip CPUs, we barely had practical transistors, we didn't have planar process circuits (invented in about '60 according to the computer history museum).
Nash believed in helping the US keep its secrets safe, and exposing the secrets of the enemy. One thing to remember when you look at Snowden's papers exposing not just domestica activities but foreign activities.
A Markov model can be represented as a finite state machine where transitions are probability driven (suppose I'm in a state 'a' and I can transition to state 'b' or 'c' where the probability of transitioning from 'a'->'b' is 30% and 'a'->'c' is 70%). The diagram on page 14 definitely looks like some sort of finite state machine but I don't know which classification it falls under. It doesn't switch states randomly so it's not a Markov chain.
A true genius of his time to us public folk, yet still behind NSA standards for security. I'm sure we will be seeing transformer-esque AI with Tesla batteries and gravity guns in the next XX (or X?) years.
[+] [-] gwern|11 years ago|reply
[+] [-] xnull2guest|11 years ago|reply
There are comments here trying to contrast Nash's work to modern cryptography, as though that's somehow fair. Nash never had access to anything remotely like a personal computer and this is just after the age - and in fact still one - where physical and mechanical devices were being used to perform encryption.
The better thing to do would be to compare (rather than contrast) his work. Look at this letter. Nash essentially predicts trapdoor functions and the P/NP gap... in a mere letter.
We were lucky to have Nash and his contributions. It's unfortunate that most of his academic work is largely unknown and that pop science focuses disproportionately on his work on equilibrium in symmetric games.
May he rest in peace.
[+] [-] kristopolous|11 years ago|reply
[+] [-] plg|11 years ago|reply
It's not like the military is not used to being precise about titles (i.e. ranks)
[+] [-] actualdc1|11 years ago|reply
Princeton considers itself a unique institution in this way. They do not accept transfer students, they do not have any professional degree programs, etc.
[+] [-] niels_olson|11 years ago|reply
[+] [-] bbulkow|11 years ago|reply
Interesting, I think, the statement that Nash's machine requires "comparatively" too much hardware. This implies the NSA did have auto-keying systems at the time that had higher security, and lower hardware burden.
Remember, in '55 we didn't have single-chip CPUs, we barely had practical transistors, we didn't have planar process circuits (invented in about '60 according to the computer history museum).
http://www.computerhistory.org/revolution/digital-logic/12/3...
Nash believed in helping the US keep its secrets safe, and exposing the secrets of the enemy. One thing to remember when you look at Snowden's papers exposing not just domestica activities but foreign activities.
[+] [-] rumdz|11 years ago|reply
[+] [-] shiro|11 years ago|reply
A few years ago I implemented it with software: http://blog.practical-scheme.net/gauche/20120715-nash-cipher...
[+] [-] hayd|11 years ago|reply
This seems a bit harsh!
[+] [-] JabavuAdams|11 years ago|reply
[+] [-] pacifist|11 years ago|reply
[+] [-] fabulist|11 years ago|reply
[+] [-] mmuelly|11 years ago|reply
[+] [-] kanan|11 years ago|reply
[+] [-] kpatterson|11 years ago|reply