US sky marshals submit to Heart of Gold randomware Californian securo-researchers are pleased to announce that America's pistol-packing undercover airline lawmen are considering the use of their cutting-edge unpredictability 'ware. The University of Southern California (USC) calls its security-operations randomness software ARMOR (Assistant for Randomized Monitoring Over …
surely the US Sky marshals are those cowboys who ride shotgun on those darned aero-me-thingies in the sky that have an aisle, or perhaps two? And this randomware <quote>can be used to make sure that security guards patrol unpredictably</quote>, up and down the one or two aisles? Right.
I take it they've not tried flicking a coin, I'm sure a quarter would be cost effective. But I suppose they'd have to nip into the toilet first to make sure no one can see them flicking the coin and exposing themselves - probably a big disadvantage.
The diffrence is that this is the US we are talking about as if they throw darts there is a chance of friendly fire or probably not even being able to hit the target at the best of times,
Mine is the one with coin in the pocket- Heads I go on flight ABC123 or Tails is flight CBA321 (I am willing to impliment this high tech and expensive system at any US airport for a bargain $2 million)
.... so it must be true.
Wow.
Now if I was a member of Al-taleban-quaida-bin-laden-bush-baby i'd currently be wetting myself. I mean what are we doing? What has the world come to? As I type this i'm genuinely questioning whether I dreamt the article... No I didn't....
Forget different approaches to randomness, for the cost of this shit you could probably put a sky marshall on every flight forthe next 20 years.
America. The race is on. Who will win first, the terrorists, or the idiots? My money's on the idiots.
What an entire nation of pond life. There's more intelligence in algae, the really is.
Oh, and Ed Blackshaw, you're a troll.
"Are you mad? Health and Safety would throw a fit!"
I doubt it. They would carefully lift a fit, move it across the room, and carefully place it down, whilst ensuring that they bend at the knees, not at the back.
Also, "fit" is probably banned as being un-PC as you might offend epileptics.
Supposedly Chinese business men (at least enough of them to warrant mentioning it) are apt to consult the I Ching or other fortune tellers when deciding whether or not (or when) to engage in a business deal. On the face of it, it sounds like a crazy thing to do, but after thinking about it several years ago, I realised that the extra randomness introduced into the financial system as a result of this custom should have the result of making their financial system more robust overall, as there'll always be some significant random element acting as a check to the purely follow-my-leader system of valuation of stocks and the attendant crowd-driven boom-and-bust element of stock markets. There are limits to the amount of "noise" that's healthy, and limits to what it can achieve (it's a refinement of the system, and not a replacement for it). But so long as not everyone goes to the same fortune teller, the result should be a more robust system.
As I said, I realised this years ago, then stumbled upon a few technical articles that seemed to back up the idea over time. Good to see that boffins in the US are waking up to the value of introducing the right kind of "noise" in an area where it can have demonstrable benefits. I do wonder about the source and quality of their RNGs though. Could we see Sky Marshals tossing yarrow stalks or rolling 3d6 and flipping pages in a choose-your-own-adventure style three-ring binder in the near future?
<-- boffinry icon as, despite appearances, there's actual science to back this idea up
The "terrists" don't need to know which flights the marshalls are on. All they need to know is the ratio of flights with marshals to flights without marshalls.
Given that there are more flights without marshalls than with marshalls, there is a good probability that their target flight will not have a marshall. And the more flights they hijack, the better their odds get of having at least one without a marshall.
Even Paris could work that one out.
The whole point about intelligence and profiling is so that you target your resources against most likely targets. The use of randomness means that you target your resources against everyone - apart from being very PC not sure that it is that logical. Maybe it is an excuse for not being effective. Or maybe they have discovered a fiendish plot by terrorists to also randomise their efforts. Are Al Qaeda at this very moment recruiting yuppie bankers or elderly English church goers as they would never be targeted as part of profiling efforts. Also, the European view was if a flight required an armed guard it should probably not take off!
Damnit, you beat me to it ... I totally agree finite improbablilty is by it's nature not even remotely random, for that you would need an infinite improbability generator ... if the US govt wants one I know this bloke with two heads who might be able to help, he's a bit unreliable though but that kinda goes with the random territory :D
Not to be too much of a spoil sport, but it does require a bit of cleverness to make it work. You can't just give each air marshal a few dice and have them roll for the next flight they will get on. Amongst other things, you need to have them home again at the end of a day's work. And you want to avoid patterns emerging that favour some flights over others due to flight scheduling quirks. So what you want is a scheduling system that provides a good spread across the flights, allows you to weight high risk routes against lower risk ones, and avoid having more than one marshal turn up on an given flight. But also one that sends any given marshal around in a part that gets them home again, whilst still maintaining the probabilistic spread of flights. That isn't so trivial. As you get to a system the size of the daily flight schedule of the entire USA, what might seem like obvious algorithms won't scale.
The press release is the usual trivialising rubbish. But it masks what is likely to be a less than trivial problem. If one insists that given enough time, every flight has a non-zero chance of getting a marshal, the problem might even reduce to a version of the travelling salesman.
I'd better preface all of this by saying that I'm not a mathematician, but I think I understand what I'm saying and I think it's a plausible interpretation. Given that, I'm not too bothered if someone proves me wrong. Anyway, one noise-related link first...
http://www.newscientist.com/article/mg16121685.100-glorious-noise.html
The gist here seems to be that adding noise to the channel can actually (counter-intuitively) increase the signal-to-noise ratio in that channel. That'd be looking at things from the Nyquist/Shannon/information-theoretic point of view. How is it relevant here? Well, assuming that going to the fortune teller introduces the right kinds of noise (white or brown, I think, but I don't know the maths here) into the system, the market as a whole should benefit by having a clearer overall picture of the true value of each stock/commodity/industry segment/whatever. This kind of noise would stand in contrast to another kind of noise that's seen in the markets--that which comes from feedback loops where good/bad news about a company causes sentiment to change which causes an increase/decrease in the stock price, which in turn causes an increase/decrease in sentiment, and so on round and round. I'm just making the argument that having the right amount of noise of the first variety can act as a counter to the second variety, leading to more accurate market valuations, and hence a more robust system overall.
Another way would be to look at things from a purely systems point of view. In that view, it's not the SNR of the channel that's important, but how susceptible the system is to being knocked out of one state of equilibrium into another (or simply lose the plot and go completely chaotic). That is, its robustness of the system would be measured by how unlikely it is to suffer dramatic readjustment. I did a quick web search along these lines and found this paper which seems to be saying exactly what I was looking for:
http://preview.tinyurl.com/5458hz
I've only read the abstract, but that seems to support what I'm saying here, at least in the case of one real-world complex system. If you understand what "limit cycles in the phase space" means (assuming we both do) then I don't think I really need to explain further why I think it supports the idea.
I can think of one other line of argument suggestive of the value of added random noise (of the appropriate kind, amount), this time from computer science... there's a wealth of information out there about using random numbers to solve or approximate various NP-hard problems. For example, using a genetic algorithm to solve a knapsack or bin-sorting problem. By way of analogy, the "solution" the system is trying to converge on would be accurate valuation of stocks, etc., while random mutation (and random selection, to a lesser degree) would take the place of throwing the yarrow stalks. I know that the systems aren't directly comparable, but I think there are enough points of similarity to throw it out there anyway.
Phew... even after the disclaimer that I'm not a mathematician, I hope I've managed to defend my original post at least passably well?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
