Schneier on Security
A blog covering security and security technology.
« Where Are All the Terrorists? |
| Outing a CIA Agent »
September 6, 2011
Optimizing Airport Security
New research: Adrian J. Lee and Sheldon H. Jacobson (2011), "The Impact of Aviation Checkpoint Queues on Optimizing Security Screening Effectiveness," Reliability Engineering & System Safety, 96 (August): 900–911.
Abstract: Passenger screening at aviation security checkpoints is a critical component in protecting airports and aircraft from terrorist threats. Recent developments in screening device technology have increased the ability to detect these threats; however, the average amount of time it takes to screen a passenger still remains a concern. This paper models the queueing process for a multi-level airport checkpoint security system, where multiple security classes are formed through subsets of specialized screening devices. An optimal static assignment policy is obtained which minimizes the steady-state expected amount of time a passenger spends in the securitysystem. Then, an optimal dynamic assignment policy is obtained through a transient analysis that balances the expected number of true alarms with the expected amount of time a passenger spends in the security system. Performance of a two-class system is compared to that of a selective security system containing primary and secondary levels of screening. The key contribution is that the resulting optimal assignment policies increase security and passenger throughput by efficiently and effectively utilizing available screening resources.
Posted on September 6, 2011 at 3:29 PM
• 21 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
Its hard to CBWFQ humans. Would be great if we could tail-drop them, but no one supports that plan.
That's. . . one heck of a lot of management-style verbiage for an abstract. Normally even this not-yet-graduated layman can understand abstracts without having to re-read them.
I suppose it means something like "use effective known techniques to sort passengers based on objective cues." But then who decides what the objective cues are? Same people who decide who gets "enhanced screening" now?
Seriously though, "increased the ability to detect threats"? Which threats? The ones that won't be used anymore because they're known and would fail? (We hope.) Right. . .
The last time I looked all practical queing systems worked on the fact that the items being qued were effectivly passive and could thus not "game the system". Which as @QueingTheory" has pointed out above does not work with humans.
Thus any sytem has to be sufficiently random that gaming it is not realistic.
The logical way to improve through put is to split people into streams on what is effectivly a "tree based" system starting with the process with minimum latency at the first choice node, and ending with the process with longest latency at the leaf choice nodes.
Obviously such a system suffers from certain problems (ie if below 5ft 8in to the left above to the right) so it needs balencing by a feed back process.
So if the first choice gate is the metal detector and only 1 in twelve people trip it but the follow up process on trip takes on average the time it takes three people to go through without triping then the process could randomly false trip on three other people to keep the que full.
This means that from the terrorists point of view they have a much greater risk of being picked up randomly than they have by trying to beat the detectors.
I guess I'm going to have to read the whole paper to see if they have a better idea.
"(ie if below 5ft 8in to the left above to the right)"
I hope you're not discriminating based on height.
Such an arbitrary measure would not only be totally ineffective, it would be trivial for the terrorists to circumvent.
All you end up with is ludicrous rules like height limits on footwear (6" heels or lower, and they must be carried in a zip-lock bag!)
Everyone knows you detect terrorists by facial hair and skin pigmentation.
Isn't starting your abstract with a tautology generally considered to be a bad idea?
I've already optimized my waiting time in airport ..um.. security queues.
"the expected number of true alarms"? Isn't that practically zero? No wait, I was thinking an actual terrorist, not just some guy with a screwdriver in his carry-on.
Does anybody know of a copy of the paper or a pre-print etc that's not stuck behind a 45USD paywall?
@ altjira: '"the expected number of true alarms"? Isn't that practically zero?'
Hmm. You might be on to something there. If P(t) is set to zero, then queueing time also drops to zero, and the only way to achieve that is to get rid of all the Potemkin nonsense.
I wonder if that's the masquerade Lee and Jacobson are attempting to play?
@Clive... this was the third hit on google, after two from schneier.com!
Although oddly the date is 2009, whereas Bruce has 2011 (the abstract looks the same but with slight punctuation changes - Bruce has a semicolon on line four, the 2009 paper starts a new sentence, for instance)
It's interesting to see that this post only has 10 comments, compared to the 60+ that usually crop up on posts that criticizes the TSA and airport security.
It would be nice if the paper was actually read and analyzed - there are a lot of people commenting here who are perfectly smart enough to do that, but it seems that the willingness to see more sides of this issue is absent.
Also, @David - Bruce didn't write that paper, so perhaps you should rethink your wording...
Wouldn't it just be easier to not allow civilians through the security lines?
Assuming David found the correct one, I read through the whole thing. Unfortunately I'm not experienced enough to check the math except superficially, at least not without taking a good deal of time about it. A few things did jump out though.
"With fixed capacities, there exists the potential for terrorists to game the system by overloading the higher levels of security with a group of high-risk operatives, thereby increasing the probability that a later-arriving, lower-risk terrorist will be insufficiently screened. However, this type of scenario can be addressed through a separate pattern detection algorithm, for example, to signal a cautionary alarm for particular sequences of passenger risk observations."
Nefarious persons might attempt to DDOS a system? Shocking!
"In practice, service times do not follow an exponential distribution, since there exists a minimum amount of time to screen a passenger or bag."
In other words, "This works mathematically, but when exposed to the real world, Murphy's Law will take over and the whole thing will revert to 2001-2011 standards of operation."
It's an idealized engineer's dream of how security ought to work, not an actual model to build today. That's not to criticize the author for writing it, just that it's only the first step in any actual reforms that may occur.
Also, Mr. Robinson, based on your comment, what do you think of the author's notion of the self-selecting process? To what extent could "enhanced screening" for this be randomized to prevent gaming the system, while at the same time trying to preserve the concept of it being a "security express line"?
@David: You presumably found the initial submission. It's common for a journal article to take 18 months from initial submission to actual publication; anything less than 12 months is warp speed....
@Thomas: "I hope you're not discriminating based on height. Such an arbitrary measure would not only be totally ineffective, it would be trivial for the terrorists to circumvent."
I think the point of Clive's comment was to simply break the line into increasingly smaller lines, not as a prerequisite to increased scrutiny.
Either way, it would not be trivial to circumvent if the left and right lines are randomized to the increased scrutiny levels.
For all practical purposes, I believe re-introducing common sense into airport security would prove a much more effective factor than spending billions on research, hardware and other services that sofar have only proven to be excellent moneymakers for some, but a major embarassment and pain in the lower backside for everyone else.
Please pardon the fact that I have dropped this link twice now, but consider its relevance as well.
If you will: http://www.activistpost.com/2011/09/...
If there is any measure of truth to this, I think it would be fair enough to give them a dose of their own snake-oil. Anyone can defend the TSA, and anyone can criticize them; the point to be reckoned with is a rapidly decaying quality of life, and the myriad forms of bureaucratic pestilence assisting it. Clearly, 9/11 has been milked so harshly, that the only thing running is the blood of liberty.
With the impending 10th anniversary of the (assisted) potential energy reduction of the World Trade Centre towers; I suddenly thought of an interesting sociological /nuclear physics question.
We, the general adherents of "the philosophy according to Bruce," are wondering, what is the half-life of security, of common sense, of personal privacy?
Given that all three are on a clear decline, has anyone tried to apply some maths (which is entirely beyond me) to the issue?
Is there a half-life? If so, when are we doomed?
@David Is there a half-life? If so, when are we doomed?
Dec 12, 2012
Reply to David (math)
A fair proposal. Never having exceeded basic algebra, I can only apply - what remains of - common sense. I would not at all be surprised if cumulative x-ray exposure and the latent (biological) effects of placing bare feet - or even bare socks - upon airport carpets, pose a threat equaling that of terrorism. Throw in backscatter vans, emotional stress, pervasive paranoia, and the National Abandonment of Reason Altogether, and you could easily get away with using a term like "decay". I fear that without a subsidized suit of critical-thinking, we are in great danger of charged p[e]ons.
I believe the date is Dec 21, 2012 (winter solstice)
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.