7 ms·

How likely is the NSA PRISM program to catch a terrorist?

181 points by johndcook 13y ago

tokenadult 13y ago
An interesting application of Bayesian reasoning to the problem of screening for low-base-rate phenomena. (And terrorist criminal activity is a lower-base-rate phenomenon in the United States, so far, than prostate cancer or other dangers that are screened for.) The mathematics, of course, is exquisitely sensitive to exactly how sensitive and specific a screening program is. (By the way, so far news reports are saying that the NSA program that is all over the front page of Hacker News, deservedly so in my opinion, is not a screening program but a data collection program, with analysis of the collected data triggered only by other kinds of law enforcement evidence-gathering. We'll see what further reporting says about that issue.)
- suredo 13y ago
  There is only one reason for collecting this data, to analyze it... and the biggest obstacle is the gathering of that data.
  - rdtsc 13y ago
    Well I guess technically there is this pesky thing getting in the way called the Constitution. But they re-defined some terms, moved some things around, decided that 'search' probably occurs when a human looks at the data (LOL, founding fathers didn't know about databases ;-). So it is not just a technical challenge, given their budgets they can solve the technical side it seems. Imagine also (and I've mentioned this before) there is no statute of limitations on historical data in case of warrant/nsl, so if warrant is every issued (and we know they are never denied) they can get _all_ your data from years and years back. Emails you sent 10 years ago to secret lovers, jokes about punching the president in the face when you were 13 years old, stuff like that. Make no mistake they will find stuff to blackmail, jail, or scare anyone into anything if they dig hard and long enough into everyone's past. I suspect more of the people involved in this will decide to do what Manning did. They see the waste, the lies, wars on terror, wars on drugs. If anything, NSA and other XYZ agencies, love to hire patriotic people, and a fraction of a percent of them maybe still are and will say "fuck it, everyone needs to know about this".
vsbuffalo 13y ago
So? 1/10,100 isn't bad at all. Suppose you have many independent tests, each creating watchlists. Suddenly you have the intersection of individuals hitting each of these, and the number of "interesting" individuals narrows. This is not a univariate problem; they're likely correlating everything to everything, and have adaptive snooping methods based on how high you are ranked a risk. Once you're high enough, field offices know about you and will keep a closer eye. It's really not a needle in a haystack (and as a privacy nerd, it pains me to say this, but I bet it is working).
- mtgx 13y ago
  I think what his data shows is that out of 10,000 terrorists, it might help catch one. Remember they've been doing this for more than a decade, and it still didn't help them with the Boston bombers. They even had the guy in custody before. Still didn't help. The abuse of surveillance power is not effective, even if it was legal/constitutional.
  - jlgreco 13y ago
    No, his calculation shows P(baddy|flag)=1/10,000. In other words, given that the system flags somebody, there is only a 1/10,000 chance that this person was correctly flagged. This is given the absurdly high P(flag|baddy) of .99 (a terrorist raises a flag 99% of the time... good luck with that.) and a P(flag|innocent person) of .01 (meaning that in the US an absolutely staggering 3.19 million people would be wrongfully flagged by this system). Then there is P(baddy)=1/1,000,000, which implies that we believe there are 300+ terrorists currently in the US that desperately need to be caught. What the hell have those 300+ terrorists been doing in the past few years? Wishing they could only find guns the the great US of A? 1/10,000,000 is a far more reasonable number, though even 1/100,000,000 is probably being generous. 1/10,000 is absolutely abysmal. When you start punching in more realistic numbers though the result becomes much worse and you rapidly approach a situation where you would have more luck flagging random US citizens as uncaught criminals than you will with using this system to identify terrorists. If you permit those sort of absurd odds, then even hiring "forensic psychics" to investigate murders in small towns starts to make sense... This sort of screening flat out does not work for something as rare as terrorists. It only works when you start bumping your P(baddy) way up (dropping P(flag|innocent person) is not practical, and raising P(flag|baddy) to even more absurdly high levels only gets you marginal improvements). Replace 'terrorist' with 'person of interest to the DEA' or 'tax cheat' and suddenly this sort of screening starts to look a hell of a lot more practical.... hmm....
mtgx 13y ago
This shouldn't even be a question. Are we going to ask how likely is that torture is going to work, too? We shouldn't allow broad surveillance of everything you ever do or say online, and we shouldn't accept torture as a means of interrogation either. This is simply deciding about the kind of society we want to be. Terrorists can be dealt with in other ways, too, and one of them would be trying to avoid creating blowback and radicalizing future generations of terrorists because of current actions.
- krichman 13y ago
  But consider that "it's dystopian and it doesn't even work" makes a strictly stronger argument. Pointing out that it won't work has better results than "those who would trade freedom for security..." on stodgy old republicans like my father (anecdata sample size 1).
- jamesaguilar 13y ago
  Maybe this makes me a bad person, but I'd ask how effective torture would be. If torturing, say, five people a year for one day each, prevented all deaths from war and terrorism for the whole year? Make that deal any day. To me the problem with torture isn't that it's "bad". It's that it doesn't serve a useful purpose and serves many counter-purposes. If it wasn't so clearly useless, I'd have to think a lot more about whether I'd support it.
  - krichman 13y ago
    Before we begin this scheme, you should have yourself and/or your family waterboarded for a few hours and see if your opinion changes. Edit: I mean, even if it were useful, is is acceptable? Probably not.
    - jamesaguilar 13y ago
      I'm not sure what my emotional reluctance to be harmed has to do with whether this (in the hypothetical world I've constructed) is a net boon for society.
    - 13y ago
craigyk 13y ago
"So you're saying there's a chance..." Haven't read it, but couldn't resist as it's an obvious conclusion.
adventured 13y ago
I wonder if the intense scrutiny these programs are coming under increases the likelihood of catching the next attempted terrorist attack by using them. (if you catch my cynical drift)
dvt 13y ago
Many of us with some (Bayesian) statistics/game-theory backgrounds are familiar with these kinds of puzzles. There are many (many) such anecdotes littered through math textbooks (HIV +/- testing comes to mind) as well as some real-life scenarios (UC Berkeley hiring practices and women comes to mind -- 1970s case). HOWEVER, let me just say that two metrics in this little puzzle are simply wrong (or at the very least unfair): P(+ | bad guy) is simply a LOT larger than 0.99 if you want to play the game correctly. The + comes from a POSITIVE outcome -- that is, a terrorist attack is thwarted and someone is thrown in jail. Some noise in the system (i.e. pizza orders) does NOT skew the positives down. After all, discerning between pizza orders and terrorist activity is part of the algorithmic process -- the automated system may flag both cases, but the buck doesn't stop there (or anywhere close). P(+ | good guy), again, suffers from the same problems as P(+ | bad guy). Unless there is evidence of 1% of the people being monitored being thrown in jail for terrorism charges, that number is a lot (lot) less than 1%. There have been plenty of bogus terrorism charges (see Guantanamo) and I think there may be something here. But if we want to play this game correctly, we need to be careful. To do this simulation, we need numbers that we will simply never have access to: e.g. how many terrorist threats did PRISM avert?
- vidarh 13y ago
  You are conflating the chance of being singled out of by PRISM with the final outcome of the process involving its use, and the benefits vs. damage done. He's not measuring, or even saying anything about, the latter. The point he is making is that to get any benefits out of PRISM, no matter how you value those benefits, you'll need to do a massive amount of further investigation to be able to identify an actual terrorist (unless you were to take drastic action; e.g. imprisoning everyone that gets "reported" - but the high number of reports to bad guys also means that becomes "impossible")
coldtea 13y ago
>How likely is the NSA PRISM program to catch a terrorist? A better question, which involves historical understanding about the issue rather than going directly to the maths and taking the framing of the thing as granted: How likely is the NSA PRISM program even cares about catching terrorists? Sure, they wouldn't say no if it did. It would even justify the program to the eyes of the public. But, thing is, similar programs have been going on for centuries in all governments. They care only about two things: spying on third countries (enemies, competitors, etc, things that effect commerce and military) and spying on their own citizens (population control, etc). Those programs have roots aeons before 9/11, and aeons before "terrorism" was any real concern of the government.
- jlgreco 13y ago
  The NSA isn't stupid. They know that screening for absurdly low incident things is an exercise in futility. I draw the same conclusion as you because of this: they don't care. ...well, not about terrorists anyway. If you are interested in flagging people that make up a much larger portion of the population then such systems are very practical. A P(flag|pot head)=.99 and P(flag|sober)=.01 would make for a distressingly effective system for example.
  - einhverfr 13y ago
    But at least they can say they are doing something. What they care about is saying that they care. It's like the satirical 2008 SwiftKids for Truth ad allegedly in support of Guiliani's Presidential Campaign: "It's time someone stood up and said 9/11... a lot"
  - s_kilk 13y ago
    I don't think it's even really about flagging anyone. The objective is more likely to be good old-fashioned careerism than anything else. These guys are in the business of "screening for terrorists" (apparently), so they are unlikely to want to see less terrorists in the world. That would put them out of a job. No, much better to expand the definition of terrorist, and expand surveilance so they have plenty of work to keep busy with.
rangibaby 13y ago
> How likely is the NSA PRISM program to catch a terrorist? Chechen Pressure Cooker Bombers: 1 NSA: 0
gordaco 13y ago
This is a very well known problem, but one that no politician on this planet seems to understand. Of course, as others have pointed out, the problem is not the likelihood of getting a bad guy. The problem is that the means are unacceptable, no matter the goal.
bane 13y ago
This is an incorrect question. Anti-terrorism is not the only thing the NSA is interested in. Better question: How likely is the NSA PRISM program to provide intelligence useful to the NSA's mission?
- flyinRyan 13y ago
  If you're going to pose that question, you better pose the other question: is NSA's mission legal/moral?
  - bane 13y ago
    I wouldn't conflate legal with moral. Those are two separate questions.
    - flyinRyan 13y ago
      I was trying to save space. That was actually two separate questions.
EliRivers 13y ago
Is that what it's for? I thought it'd be a big retroactive database; identify someone we want to know about now, and then look back through the last year's records to see what she's been doing.
flyinRyan 13y ago
Well, it's been running for a while, how many has it caught? Oh, zero? Yea, that's what I expected.

181 points by johndcook 13y ago

An interesting application of Bayesian reasoning to the problem of screening for low-base-rate phenomena. (And terrorist criminal activity is a lower-base-rate phenomenon in the United States, so far, than prostate cancer or other dangers that are screened for.) The mathematics, of course, is exquisitely sensitive to exactly how sensitive and specific a screening program is. (By the way, so far news reports are saying that the NSA program that is all over the front page of Hacker News, deservedly so in my opinion, is not a screening program but a data collection program, with analysis of the collected data triggered only by other kinds of law enforcement evidence-gathering. We'll see what further reporting says about that issue.)

suredo 13y ago
There is only one reason for collecting this data, to analyze it... and the biggest obstacle is the gathering of that data.
- rdtsc 13y ago
  Well I guess technically there is this pesky thing getting in the way called the Constitution. But they re-defined some terms, moved some things around, decided that 'search' probably occurs when a human looks at the data (LOL, founding fathers didn't know about databases ;-). So it is not just a technical challenge, given their budgets they can solve the technical side it seems. Imagine also (and I've mentioned this before) there is no statute of limitations on historical data in case of warrant/nsl, so if warrant is every issued (and we know they are never denied) they can get _all_ your data from years and years back. Emails you sent 10 years ago to secret lovers, jokes about punching the president in the face when you were 13 years old, stuff like that. Make no mistake they will find stuff to blackmail, jail, or scare anyone into anything if they dig hard and long enough into everyone's past. I suspect more of the people involved in this will decide to do what Manning did. They see the waste, the lies, wars on terror, wars on drugs. If anything, NSA and other XYZ agencies, love to hire patriotic people, and a fraction of a percent of them maybe still are and will say "fuck it, everyone needs to know about this".

vsbuffalo 13y ago

So? 1/10,100 isn't bad at all. Suppose you have many independent tests, each creating watchlists. Suddenly you have the intersection of individuals hitting each of these, and the number of "interesting" individuals narrows. This is not a univariate problem; they're likely correlating everything to everything, and have adaptive snooping methods based on how high you are ranked a risk. Once you're high enough, field offices know about you and will keep a closer eye. It's really not a needle in a haystack (and as a privacy nerd, it pains me to say this, but I bet it is working).

mtgx 13y ago
I think what his data shows is that out of 10,000 terrorists, it might help catch one. Remember they've been doing this for more than a decade, and it still didn't help them with the Boston bombers. They even had the guy in custody before. Still didn't help. The abuse of surveillance power is not effective, even if it was legal/constitutional.
- jlgreco 13y ago
  No, his calculation shows P(baddy|flag)=1/10,000. In other words, given that the system flags somebody, there is only a 1/10,000 chance that this person was correctly flagged. This is given the absurdly high P(flag|baddy) of .99 (a terrorist raises a flag 99% of the time... good luck with that.) and a P(flag|innocent person) of .01 (meaning that in the US an absolutely staggering 3.19 million people would be wrongfully flagged by this system). Then there is P(baddy)=1/1,000,000, which implies that we believe there are 300+ terrorists currently in the US that desperately need to be caught. What the hell have those 300+ terrorists been doing in the past few years? Wishing they could only find guns the the great US of A? 1/10,000,000 is a far more reasonable number, though even 1/100,000,000 is probably being generous. 1/10,000 is absolutely abysmal. When you start punching in more realistic numbers though the result becomes much worse and you rapidly approach a situation where you would have more luck flagging random US citizens as uncaught criminals than you will with using this system to identify terrorists. If you permit those sort of absurd odds, then even hiring "forensic psychics" to investigate murders in small towns starts to make sense... This sort of screening flat out does not work for something as rare as terrorists. It only works when you start bumping your P(baddy) way up (dropping P(flag|innocent person) is not practical, and raising P(flag|baddy) to even more absurdly high levels only gets you marginal improvements). Replace 'terrorist' with 'person of interest to the DEA' or 'tax cheat' and suddenly this sort of screening starts to look a hell of a lot more practical.... hmm....

mtgx 13y ago

This shouldn't even be a question. Are we going to ask how likely is that torture is going to work, too? We shouldn't allow broad surveillance of everything you ever do or say online, and we shouldn't accept torture as a means of interrogation either. This is simply deciding about the kind of society we want to be. Terrorists can be dealt with in other ways, too, and one of them would be trying to avoid creating blowback and radicalizing future generations of terrorists because of current actions.

krichman 13y ago
But consider that "it's dystopian and it doesn't even work" makes a strictly stronger argument. Pointing out that it won't work has better results than "those who would trade freedom for security..." on stodgy old republicans like my father (anecdata sample size 1).
jamesaguilar 13y ago
Maybe this makes me a bad person, but I'd ask how effective torture would be. If torturing, say, five people a year for one day each, prevented all deaths from war and terrorism for the whole year? Make that deal any day. To me the problem with torture isn't that it's "bad". It's that it doesn't serve a useful purpose and serves many counter-purposes. If it wasn't so clearly useless, I'd have to think a lot more about whether I'd support it.
- krichman 13y ago
  Before we begin this scheme, you should have yourself and/or your family waterboarded for a few hours and see if your opinion changes. Edit: I mean, even if it were useful, is is acceptable? Probably not.
  - jamesaguilar 13y ago
    I'm not sure what my emotional reluctance to be harmed has to do with whether this (in the hypothetical world I've constructed) is a net boon for society.
  - 13y ago

craigyk 13y ago

"So you're saying there's a chance..." Haven't read it, but couldn't resist as it's an obvious conclusion.

adventured 13y ago

I wonder if the intense scrutiny these programs are coming under increases the likelihood of catching the next attempted terrorist attack by using them. (if you catch my cynical drift)

dvt 13y ago

Many of us with some (Bayesian) statistics/game-theory backgrounds are familiar with these kinds of puzzles. There are many (many) such anecdotes littered through math textbooks (HIV +/- testing comes to mind) as well as some real-life scenarios (UC Berkeley hiring practices and women comes to mind -- 1970s case). HOWEVER, let me just say that two metrics in this little puzzle are simply wrong (or at the very least unfair): P(+ | bad guy) is simply a LOT larger than 0.99 if you want to play the game correctly. The + comes from a POSITIVE outcome -- that is, a terrorist attack is thwarted and someone is thrown in jail. Some noise in the system (i.e. pizza orders) does NOT skew the positives down. After all, discerning between pizza orders and terrorist activity is part of the algorithmic process -- the automated system may flag both cases, but the buck doesn't stop there (or anywhere close). P(+ | good guy), again, suffers from the same problems as P(+ | bad guy). Unless there is evidence of 1% of the people being monitored being thrown in jail for terrorism charges, that number is a lot (lot) less than 1%. There have been plenty of bogus terrorism charges (see Guantanamo) and I think there may be something here. But if we want to play this game correctly, we need to be careful. To do this simulation, we need numbers that we will simply never have access to: e.g. how many terrorist threats did PRISM avert?

vidarh 13y ago
You are conflating the chance of being singled out of by PRISM with the final outcome of the process involving its use, and the benefits vs. damage done. He's not measuring, or even saying anything about, the latter. The point he is making is that to get any benefits out of PRISM, no matter how you value those benefits, you'll need to do a massive amount of further investigation to be able to identify an actual terrorist (unless you were to take drastic action; e.g. imprisoning everyone that gets "reported" - but the high number of reports to bad guys also means that becomes "impossible")

coldtea 13y ago

>How likely is the NSA PRISM program to catch a terrorist? A better question, which involves historical understanding about the issue rather than going directly to the maths and taking the framing of the thing as granted: How likely is the NSA PRISM program even cares about catching terrorists? Sure, they wouldn't say no if it did. It would even justify the program to the eyes of the public. But, thing is, similar programs have been going on for centuries in all governments. They care only about two things: spying on third countries (enemies, competitors, etc, things that effect commerce and military) and spying on their own citizens (population control, etc). Those programs have roots aeons before 9/11, and aeons before "terrorism" was any real concern of the government.

jlgreco 13y ago
The NSA isn't stupid. They know that screening for absurdly low incident things is an exercise in futility. I draw the same conclusion as you because of this: they don't care. ...well, not about terrorists anyway. If you are interested in flagging people that make up a much larger portion of the population then such systems are very practical. A P(flag|pot head)=.99 and P(flag|sober)=.01 would make for a distressingly effective system for example.
- einhverfr 13y ago
  But at least they can say they are doing something. What they care about is saying that they care. It's like the satirical 2008 SwiftKids for Truth ad allegedly in support of Guiliani's Presidential Campaign: "It's time someone stood up and said 9/11... a lot"
- s_kilk 13y ago
  I don't think it's even really about flagging anyone. The objective is more likely to be good old-fashioned careerism than anything else. These guys are in the business of "screening for terrorists" (apparently), so they are unlikely to want to see less terrorists in the world. That would put them out of a job. No, much better to expand the definition of terrorist, and expand surveilance so they have plenty of work to keep busy with.

rangibaby 13y ago

> How likely is the NSA PRISM program to catch a terrorist? Chechen Pressure Cooker Bombers: 1 NSA: 0

gordaco 13y ago

This is a very well known problem, but one that no politician on this planet seems to understand. Of course, as others have pointed out, the problem is not the likelihood of getting a bad guy. The problem is that the means are unacceptable, no matter the goal.

bane 13y ago

This is an incorrect question. Anti-terrorism is not the only thing the NSA is interested in. Better question: How likely is the NSA PRISM program to provide intelligence useful to the NSA's mission?

flyinRyan 13y ago
If you're going to pose that question, you better pose the other question: is NSA's mission legal/moral?
- bane 13y ago
  I wouldn't conflate legal with moral. Those are two separate questions.
  - flyinRyan 13y ago
    I was trying to save space. That was actually two separate questions.

EliRivers 13y ago

Is that what it's for? I thought it'd be a big retroactive database; identify someone we want to know about now, and then look back through the last year's records to see what she's been doing.

flyinRyan 13y ago

Well, it's been running for a while, how many has it caught? Oh, zero? Yea, that's what I expected.