" /> Screenshot: A Weblog: January 2008 Archives

« December 2007 | Main | February 2008 »

January 21, 2008

It's 11 o'clock. Do you know where your data is?

I've been keeping track of interesting stories about security over the past couple of months for my intersession course, mostly ones that I have found through Slashdot, BoingBoing and/or Digg. As part of the process of selecting which ones will make it into the final week of the course and which ones will not, I thought I would put the whole list here, mostly without comment. If you notice anything that seems particularly interesting (especially if you happen to be in my course!) let me know and it will probably move up my list of things to discuss.

January 9, 2008

Avoiding a pun about "crashing"....

I've been saving up news articles about security vulnerabilities for my cyberattacks class, but I'm not quite sure where to fit in a discussion of potential vulnerabilities in Boeing's New 787. On the crazy-cool side, the plane is going to have internet connectivity in the cabin for passengers. On the crazy-stupid side, the passenger's network is connected to the cockpit network. Solutions are being discussed, but they do not seem to include just keeping the two networks physically separate. But software solutions can, and probably will, have holes, and Boeing is treating this as a software-debugging problem. I can't imagine what the justification would be for wanting the networks to be connected. I am a big proponent of the "if it is absolutely vital, keep it unplugged from any network" school of security. Or, frankly, if you can't do it safely, I'll get by without internet access on my plane flight....

January 2, 2008

Or you could just let Pennsylvania go first....

I was having a discussion over dinner about the problems with current presidential primary process, particularly the scheduling of them - yeah, I know, it's a controversial stance! Pretty much everybody you talk to has an idea for what could make the current scheduling better, and we were arguing the merits of various hypothetical plans when somebody observed that perhaps people with more expertise and who had actually analyzed the relevant data had looked at this question. So it was home to Wikipedia and their US Presidential Primary page, and the also good FairVote page on Presidential Primaries.

The major variations seem to involve either (1) group primaries starting with small states, and then working up to larger states towards the end of the process, (2) ordering the primaries to start with a random sampling of primaries but with structure imposed to start with "easy" primaries and work up to the larger, more expensive ones, (3) working through regions of the country in turn, or (4) pulling one state from each of a set of regions for each of a set of primary dates. FairVote has nice details on how each of these work with sample breakdowns/schedules.

The cynic in me thinks it likely, though, that any of these plans is going to lead towards a bias towards particular groups/regions and against others, and that saavy analysts will be able to work out which these are and the constituency with the best lobbying power is going to win (if anything ends up changing). To me, this calls out for a different plan (yep, despite what I said about listening to people who actually know what they are talking about, I'm going to throw in my ignorant two cents...) based on pure randomness. Let's pick a set of primary dates, and then randomly order the states among those dates. In order to prevent a state from being consistently devalued by falling late in the process, if you are in the last quarter of the primaries in one cycle, you are guaranteed to be in the first half of the primaries in the next cycle.

Sure, in any given year, you could have a bad outcome - small states could get a disproportionate say, primaries could be located such that poorer candidates have a harder time competing, etc. But you would avoid systematic biasing and considering the long-range trends of presidential elections, these concerns ought to even out. Otherwise, the debate seems to focus on whether particular goals (giving larger, urban states more say, making campaigning easier on fringe candidates, etc.) actually is desirable or not. And as I like to remind my students when looking at various AI systems, you always want to ask yourself if your highly engineered system beats random chance....