More Details of the NSA's Social Network Analysis

mrogers writes "USA Today has a story describing how the NSA looks for suspicious calling patterns in the huge volumes of traffic data it collects. "Templates" such as a call from overseas followed by a flurry of domestic calls are used to identify leads, which are forwarded to the FBI for investigation. There have been complaints that low-quality leads are drawing agents away from other cases, and similar pattern-matching approaches have been found wanting in the past. Can data mining identify terrorists?"

Comments

Beside the point.

[TripMaster Monkey]

If this wholesale data mining works, then the government will tout this success as justification for its acts. If it doesn't work, the government will complain that we're not letting them do enough to ensure our safety, and use the failure to justify even more outrageous violations of our privacy.

Whether it works or not, however, is beside the point. The point is: is it legal? Enough people have maintained that it is not to warrant a serious investigation into the matter.

My grandmother may be a terrrorist.

[JesseL]

She's always getting calls from various places and then making a flurry of more local calls. She uses code phrases like "your cousin's baby was born last night and it's a boy", or "Great Aunt Zelda had a stroke but they say she's going to be okay".

Raise it to orange

[Trails]

"Hey Akbar, just calling to let you know Mohamed and Alimah just had a healthy baby boy!"

"Oh great, I'll let the family over here know!"

*meanwhile, in the basement of a bunker somewhere*

"My God! It's nine eleven times ten thousand! Nine million one hundred and ten thousand!"

Simple answer? Kinda

[Red Flayer]

"Can data mining identify terrorists?"

No. It can identify people who have calling patterns associated with terrorist activity, regardless of whether they are a terrorist or not.

Note that these calling patterns cannot be used to associate that person with a committed or planned crime in the normal data mining scenario.

Data mining is unreasonable search.

Now, I have no problem if they've got evidence of a crime or plan of a crime, and use known information to deduce who might else be involved. That's investigative work.

Data mining is speculative work, not investigative, so regardless of whether it *can* be used for speculative 'research' into the activity of American citizens, it *shouldn't* be.

What ever happened to....

[beheaderaswp]

What ever happened to "Live free or die", "Give me liberty or give me death", or "Those who are willing to sacrifice their basic liberties to assure their security deserve neither."?

Those quotes are not just platitudes... they are *good ideas*.

Keep the canned patriotism, give me my rights, and I'll just take my chances.

Re:Quick, Look the Other Way!

[dhasenan]

Because in order for your vote to count, it has to agree with a large number of other votes. If we got a libertarian for President--say, Michael Badnarik--then the NSA would have to hide its spying from the President, as well. But for any national candidate to succeed, they need media coverage. For some reason, Ralph Nader, who was only on the ballot in 36 states, got far more coverage than Badnarik, who was on the ballot in (I believe) 49 states. Why? Because Nader couldn't have won, so the media could safely involve him.

So, your choices for every election are between media coalitions. Which generally means that each of the major US parties supports slightly differing sections of the economy--service sector for the Democrats, production for the Republicans. That's the major difference.

Now, armed resistance is ridiculous when the government has billions of dollars of military equipment. And other technological countermeasures will likely prove ineffective in a short period of time.

Disarm them.

[babbling]

The most effective way of stopping terrorists is taking away their cause. Believe it or not, terrorists don't blow up hundreds of people as well as themselves because they "hate freedom" or any of that rubbish.

Pipe Dreams

[Khammurabi]

Can data mining identify terrorists?

Not really. Computers are good at recognizing patterns only when there is a large repository of data to "train" the computer with. For example, neural networks are often better at recognizing patterns than if a person were to program a set of rules into a system. Man-made rules are often incomplete or lack the depth that a computer can bring to the table. A good example of this is Google Translate, which is considered one of the better translation programs and is essentially an advanced neural net that was fed a huge wad of data to train from.

America's data set on terrorism is in the single digits, and the data they do have is only partially complete. This means the only system that can be programmed is a set of user-created rules that "flag" questionable behavior. The solution is a poor one and will only improve our chances at detection by a fraction of a percent. (Seems a huge price to pay for privacy trampling to me.)

In order to detect terrorism on American soil effectively, we'd need a larger data set. Otherwise we're just attempting to reverse engineer a process that essentially defines itself as dynamic enough to avoid detection. We'd need a frequent source of terrorism that we could derive models and nets off of. The immediate source that comes to mind is Iraq. If I were in charge of the NSA program, I think the best course of action would be to harness the call-traffic (satellite and domestic), email activity and other "data" that precedes suicide bombers (or other known acts of terrorism) in Iraq. Using this data you could train a system to recognize similarities in America. Short of that, anything the NSA is trying is a crap shoot.

No. Freeing up lines of communication, preparing quick and actionable responses to warnings, and better general population awareness are probably more effective than grabbing a billion pieces of data and sifting through it for answers. It's impossible for a human to know what to look for, and until the NSA comes clean in what it's actualy doing, there's no justification for stomping out the few freedoms we still have. There are better alternatives out there that can be done with the help of the community and still preserve the integrity of our privacy.

Re:The strength of weak links...

[noewun]

Was about to say the same thing. Traffic pattern analysis doesn't work at all for sleeper cells, like the people who carried out the 9/11 attacks. Sleeper cells, by definition, tend to be quiet for long periods of time with only intermittent contact between members and any organizing force. To someone looking at traffic pattern analysis, this will look no different from me talking to my cousin in Atlanta or my uncle in DC, which we do once in a long while. Analysis of the 9/11 hijackers would've shown normal, suburban usage.

The trend in terrorism lately is decentralization: the guys who carried out the Madrid train bombings were home-grown, were not known terrorists, and were not previoiusly involved in any high level attacks or meetings. They didn't show up on anyone's radar precisely because they didn't fit any profile, nor would they be found with traffic pattern analysis. Add to this the recent news that the AQ higher ups have ceased using satellite or cel phones and you have the basic problem with asymetrical warfare, one which the White House and DoD refuse to learn: you can't fight a guy wearing a suicide vest with satellites and computers, and you can't find a loosely organized, ad hoc group of people by looking for organized cells. The top down model of terrorism is dead, and it seems to be the only thing we're still looking for.

What we need, and what the White House and DoD are steadfastly refusing to develop, is old-fashioned HUMINT, human intelligence. We need speakers of Arab in all of the various dialects, we need people schooled in Middle Eastern politics, history, religion and socities, and we need to get people with Middle Eastern backgrounds into the intelligence services and up the command chain. One of the reasons the CIA was as efficient as it was in the 60s and 70s was the large number of working agents from countries in which they were working. Gust Avrakotos was such an effective agent in Greece and elsewhere because he spoke the native languages and knew the local customs. He wasn't viewing the space by satellite from DC. He was in the mix.

Here endeth the rant.