Skip to main content

View Diary: More on Research 2000 (474 comments)

Comment Preferences

  •  no really it's not like drug testing (0+ / 0-)

    and computers can use real physical non-algorithmic sources of randomness, similar to flipping coins.  That book looks interesting but I think you must be misinterpreting it a little.  There's just no way to distinguish computer-generated randomness (even pseudo-randomness, given a small amount of real randomness to seed the generator with) from any other kind.  A lot of computer security (like encryption) depends on that principle, so techies pay close attention to it.

    •  I'm no expert (0+ / 0-)

      but I don't see how encryption depends on randomness - thought that was more about the unique prime factorization of encryption keys, which is decidedly non-random (or is the pseudo-randomness you're referring to that the keys have to be 'randomly' selected?)

      if you're still around would appreciate any insight you have on that

      You couldn't load a pistol with dormitive virtue and shoot it into a breakfast-roll - CS Pierce

      by Mr Raymond Luxury Yacht on Tue Jun 29, 2010 at 11:36:55 PM PDT

      [ Parent ]

      •  randomness (1+ / 0-)
        Recommended by:
        Mr Raymond Luxury Yacht

        The thing with the prime numbers is that if I can guess your primes, I can read your messages.  And if you generated them with some deterministic program, I can recreate them the same way.  So the right way to generate them is at random, by flipping coins or doing the digital equivalent.  That's one place randomness comes in.

        What I was getting at though was a bit different.  Say you encrypt a secret message and send it to your friend, and I intercept it and I want to read it but can't, since I don't know the secret key (that can be prime numbers, a password, or whatever).  In the old days that would have been considered good encryption, if I can't read the message.  These days we'd say, even if I can't read the message, maybe it reveals enough info to let me statistically estimate the ratio of vowels to consonants.  That in turn informs me whether the message is written in English or Russian, which may be enough for me to bust your spy ring, even if I can't tell what the message actually says.  So we say encryption is no good if it reveals anything at all about the message.  And what is the standard for not revealing anything?

        Suppose you send a second "message", which actually has no password and no plaintext.  It's just purely random (from coin flips).  I intercept both messages and I know that one of them is encrypted (but meaningful, at least to you) and one is random (i.e. meaningless), but I don't know which is which.  The standard of a good encryption method is that I can't tell the two apart in any feasible way (doing a billion year brute force guessing attack doesn't count as "feasible").  I shouldn't be able to distinguish a meaningful message from a random one.  If I can do that, your code is considered no good, even if it doesn't leak the actual contents.  

        So the security community spends a lot of time concocting ways to make meaningful messages indistinguishable from random ones.  They are quite successful at this.  And somebody trying to fake polls can use similar methods to generate result sets that can't be distinguished from real ones.

        •  Gotcha, I see where you're coming from (0+ / 0-)

          You hit it on the head - the security community spends all kinds of time and money to make encrypted messages look random, resources that Ali's polling outfit probably doesn't have.

          Be fascinating to see what happens when (if?...nah, when) quantum computing becomes a reality and factoring large numbers takes seconds instead of billions of years.....

          You couldn't load a pistol with dormitive virtue and shoot it into a breakfast-roll - CS Pierce

          by Mr Raymond Luxury Yacht on Wed Jun 30, 2010 at 12:17:54 PM PDT

          [ Parent ]

          •  well, the work they do is in figuring out how (0+ / 0-)

            to make the messages look random.  Once the big brains (mostly in academia) have figured out and published the methods, smaller brains like mine can read the publications and follow the instructions.  Or even simpler, there's tons of programs out there that anyone can simply download and use for free.  I'm pretty convinced that anyone with some basic mathematical knowledge (well below the level of someone who works as a statistician at a polling company) could cook a poll in a way that even an expert couldn't tell what happened.  If R2K didn't have the skill to do that, it probably speaks poorly of the reliability of even their legitimate polling.

            •  I think what we're going to learn (0+ / 0-)

              is that R2K wasn't getting enough phone pick-ups from their outreach efforts, and didn't have the cash flow to fund more dialers.  Fraud that crept in from desperation rather than was coordinated from the outset.  Just my guess though.  But it would explain the shitty job they ended up doing on it.

              You couldn't load a pistol with dormitive virtue and shoot it into a breakfast-roll - CS Pierce

              by Mr Raymond Luxury Yacht on Wed Jun 30, 2010 at 02:46:20 PM PDT

              [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site