NSA Spying Doesn’t Work to Prevent Terrorism
William Binney knows as much about spying as anyone alive.
Binney – a 32-year National Security Agency veteran – is the former head of the NSA’s global digital data gathering program, and a very highly-regarded cryptographer.
Binney told Daily Caller yesterday that the spying “dragnet” being carried out by the government is useless:
Daily Caller: There’s been some talk about the authorities having a recording of a phone call Tamerlan Tsarnaev had with his wife. That would be something before the bombing?
Binney: Before the bombing, yes. [This information comes from former FBI counterrorism agent Tim Clemente.]
Daily Caller: Then how would they have that audio?
Binney: Because the NSA recorded it.
Daily Caller: But apparently the Russians tipped off the FBI, which then did a cursory interview and cleared him. So how were they recording him?
Binney: Because the Russians gave a warning for him as a target. Once you’re on a list, they start recording everything. That’s what I’m saying.
Daily Caller: So why didn’t they prevent the bombing?
Binney: Once you’ve recorded something, that doesn’t mean they have it transcribed. It depends on what they transcribe and what they do with the transcription.
Daily Caller: So it seems logical to ask: Why do we need all of this new data collection when they’re not following up obvious leads, such as an intelligence agency calling and saying you need to be aware of this particular terrorist?
Binney: It’s sensible to ask, but that’s exactly what they’re doing. They’re making themselves dysfunctional by collecting all of this data. They’ve got so much collection capability but they can’t do everything.
Daily Caller: So what are they doing with all of this information? If they can’t stop the Boston marathon bombing, what are they doing with it?
Binney: Well again, they’re putting an extra burden on all of their analysts. It’s not something that’s going to help them; it’s something that’s burdensome. There are ways to do the analysis properly, but they don’t really want the solution because if they got it, they wouldn’t be able to keep demanding the money to solve it. I call it their business statement, “Keep the problems going so the money keeps flowing.” It’s all about contracts and money.
Daily Caller: But isn’t data collection getting easier and processing speeds getting faster and data collection cheaper? Isn’t the falling price one of the reasons they can collect data at this massive level?
Binney: Yes, but that’s not the issue. The issue is, can you figure out what’s important in it? And figure out the intentions and capabilities of the people you’re monitoring? And they are in no way prepared to do that, because that takes analysis. That’s what the big data initiative was all about out of the White House last year. It was to try to get algorithms and figure out what’s important and tell the people what’s important so that they can find things. The probability of them finding what’s really there is low.
Similarly, Fortune notes that the NSA’s “big data” strategy is ineffective:
The evidence for big data is scant at best. To date, large fields of data have generated meaningful insights at times, but not on the scale many have promised. This disappointment has been documented in the Wall Street Journal, Information Week, and SmartData Collective.
According to my firm’s research, local farmers in India with tiny fields frequently outperform — in productivity and sustainability — a predictive global model developed by one of the world’s leading agrochemical companies. Why? Because they develop unique planting, fertilizing, or harvesting practices linked to the uniqueness of their soil, their weather pattern, or the rare utilization of some compost. There is more to learn from a local Indian outlier than from building a giant multivariate yield prediction model of all farms in the world. The same is true for terrorism. Don’t look for a needle in a giant haystack. Find one needle in a small clump of hay and see whether similar clumps of hay also contain needles.
You need local knowledge to glean insights from any data. I once ran a data-mining project with Wal-Mart (WMT) where we tried to figure out sales patterns in New England. One of the questions was, “Why are our gun sales lower in Massachusetts than in other states, even accounting for the liberal bias of the state?” The answer: There were city ordinances prohibiting the sale of guns in many towns. I still remember the disappointed look of my client when he realized the answer had come from a few phone calls to store managers rather than from a multivariate regression model.
So, please, Mr. President, stop building your giant database in the sky and quit hoping that algorithm experts will generate a terrorist prevention strategy from that data. Instead, rely on your people in the field to detect suspicious local patterns of behavior, communication, or spending, then aggregate data for the folks involved and let your data hounds loose on these focused samples. You and I will both sleep better. And I won’t have to worry about who is lurking in the shadows of my business or bedroom.
Likewise, Nassim Taleb writes:
Big data may mean more information, but it also means more false information.
Because of excess data as compared to real signals, someone looking at history from the vantage point of a library will necessarily find many more spurious relationships than one who sees matters in the making; he will be duped by more epiphenomena. Even experiments can be marred with bias, especially when researchers hide failed attempts or formulate a hypothesis after the results — thus fitting the hypothesis to the experiment (though the bias is smaller there).