Making Hay of the NSA
My earlier post about “No Such Agency” drew a good amount of attention, including both kudos and criticism. I heartily welcome the criticism as the sign that we’re having a valuable debate. We never would have known to have this debate without four brave citizens coming forward in spite of great personal consequences to themselves and their families.
I want to respond to the most common thread espoused by critics. I call it the data mining argument.
The most commonly quoted line of my original opinion piece was this metaphor: is the solution to finding a needle in a haystack— adding more hay?
I am referring, of course, to the vacuuming of all information, constitutional and otherwise, in order to then mine that information for clues. The alternative would be targeted collection of data from warranted searches, both foreign and domestic.
The critics argued that “adding more hay” by collecting more unwarranted data from American citizens would be a way to provide training data to discern a signal in the noise. They suggest one could mine this data to discover terrorists. There is one classic software engineering problem with this theory— garbage in, garbage out.
“Garbage in, garbage out” means that if you input bad data into a computer algorithm, you are likely to get bad output. I agree with computer security expert, Bruce Schneier. Adding more, low-quality training data cannot work miracles.
If our goal is to discern terrorists, then we have a problem. High-level terrorists do not communicate electronically. For example, the wealthy financier Osama Bin Laden had no phone lines and no connection to the Internet for many years. This is why the top intelligence agencies in the world could not find him for many years. When they do attempt to communicate through intermediaries, they attempt to bury their signal in the noise.
“Garbage in” tends to result in false negatives and false positives.
False negatives are when we falsely identify a guilty communication as innocent. This results in worse training data and missed opportunities to catch the bad guys. This most often happens because the bad guys have figured tricks to elude automatic detection, similar to methods spammers employ to dodge predictive spam detection models.
False positives are when we falsely identify an innocent communication as guilty. These are an immense problem because they result in wasted intelligence resources and the kidnapping, detention, and potential torture of innocents.
False positives can also badly poison the training data. Innocents will lie to investigators to make torture or “enhanced interrogation” stop. At the height of the war in Iraq, we discovered we likely had a significant number of innocent prisoners in Guantanamo. Any training data that assumed they and their communications were guilty would also have poisoned training data and caused more mistakes later on.
Perhaps the greatest problem with the data mining argument is that it assumes the primary goal of intelligence agencies is to catch terrorists. Outside of Israel, this is false.
In America, terrorism is not an existential threat. It is a real threat and a manageable one.
Rash actions impede our ability to catch bad guys. Enhanced assassination programs and enhanced interrogation do not likely enhance our our ability to stop terrorism. Our collateral damage instead harms us and our image in the world.
Adding more bad training data by including the private lives of American citizens does not seem a worthy sacrifice for a problem that causes far fewer deaths and injuries than bathtubs.
To those who call the bathtub metaphor an unfair comparison, consider this. 9/11 broke the bell curve of modern terrorism in America by several degrees of magnitude. It is a feat not likely to be repeated.
9/11 only occurred because Bin Laden possessed a great Saudi oil fortune that he spent on attacking us. The greatest blow dealt to terrorism was not the “Patriot” Act or PRISM, but cutting up his credit cards and closing his bank accounts. Now only state-sponsored terror and blowback from our own well-funded spy operations offers much real threat to us. (Note: Bin Laden himself is a classic example of “blowback,” as America hired him and paid him to fight the Soviets in Afghanistan.)
Contrary to popular perceptions, most of the Muslim world is either supportive or neutral to the United States. This is why so many come to live here.
9/11 was hardly the first time Americans feared irregular warriors in their daily lives. It is particularly telling that Bin Laden was codenamed “Geronimo” by our military, a glaring insult to our Native American population.
Are we still dreading being scalped today? Or is “Cowboys and Indians” merely a racially-insensitive game that trivializes a contentious period of history when Native Americans were responding to being terrorized on the American frontier?
At one point in our history, everything being said against ordinary Muslims was also being said against Native Americans.
The threat posed by Muslim extremists is not something to fear in our daily lives. Fear and resentment harmed the effort to resist terrorism in Israel and it will do the same if we let it define us in America. We are better than this.
Again, I believe the National Security Agency has provided a valuable service to our nation in years past. I just want to know that our Constitution is followed. Someone honest must watch the watchers.
