The revelation that the National Security Agency has been wiretapping communications crossing the U.S. border (and possibly within the U.S.), without warrants, has started many angry conversations across the country, and rightly so. Here is an issue that challenges our most basic conception of the purposes of government and its relation to citizens.
Today I am starting a series of posts about this issue. Most discussions of the wiretap program focus on two questions: (1) Is the program legal? and (2) Regardless of its legality, does the program, as currently executed, serve our national interest (bearing in mind the national interest in both national security and citizens’ privacy)? These questions are surely important, but I want to set them aside here. I’m setting aside the legal question because it’s outside my expertise. I’m setting aside any evaluation of the current program for two reasons. First, we don’t know the exact scope of the current wiretap program. Second, most people – on both sides – think the second question is an easy one, and easy questions lead to boring conversations.
I want to focus instead on the more basic questions of what the extent of national security wiretapping should be, and why. The why question is especially important.
The first thing to realize is that this is not your parents’ wiretap debate. Though the use (and sometimes misuse) of wiretapping has long been a contentious issue, the terms of the debate have changed. I’m not referring here to the claim that 9/11 changed everything. What I mean is that wiretapping technology has changed in ways that ought to reframe the debate.
Two technology changes are important. The first is the dramatic drop in the cost of storage, making it economical to record vast amounts of communications traffic. The second technology change is the use of computer algorithms to analyze intercepted communications. Traditionally, a wiretap would be heard (or read) immediately by a person, or recorded for later listening by a person. Today computer algorithms can sift through intercepted communications, looking for sophisticated patterns, and can select certain items to be recorded or heard by a person.
Both changes are driven by Moore’s Law, the rule of thumb that the capability of digital technologies doubles every eighteen months or, equivalently, improves by a factor of 100 every ten years. This means that in 2016 government will be able to store 100 times more intercepted messages, and will be able to devote 100 times more computing capability to its analysis algorithms, compared to today. If the new world of wiretapping has not entirely arrived, it will be here before long.
So government will have greater eavesdropping capabilities and, more interestingly, it will have different capabilities. How should we respond? Surely it is not right simply to let government do whatever it wants – this has never been our policy. Nor can it be right to let government do no wiretapping at all – this has not been our policy either. What we need to understand is where to draw the line, and what kind of oversight and safeguards we need to keep our government near the line we have drawn. I hope that the next several posts can shed some small amount of light on these questions.
It would explain the Fermi paradox. On the other hand, there aren’t very many alternatives known to physicists for long range comms. EM is efficient in the transparent vacuum of space, as long as you stay away from wavelengths readily absorbed by hydrogen gas; it doesn’t take megaengineering to make a signal detectable over a respectable distance. The other obvious candidates are gravitational radiation, neutrinos, and neutralinos. The first requires megaengineering to emit and fancy equipment to detect, without much better penetrating power. The next is even harder to detect and requires careful modulation of nuclear reactions to use for signaling. The last, well, forgeddaboudit unless you can routinely reach grand unification energy in your transmitter and expect your pen pals to do so in their receivers as well.
Communication by modulation of EMR in the grand scale of things may be as crude and ineffective as smoke signals bringing the Arapahoe to the attention of smoke signallers on the great wall of China in the Ming dynasty.
We may discover a far superior interstellar communications mechanism, and radio would become obsolete pretty quickly.
One situation in which SETI could succeed is where a civilization purposely broadcasts in order to say “Hey, anyone out there?”; such a signal would be purposely easy to decode and sent either omnidirectionally or targeted at various likely candidate stars, possibly including ours.
Our own civilization has generated at least two such signals, using as I recall the Arecibo radio telescope as transmitter. So far, no discernible reply.
If they come knocking though, it’s a toss-up whether there’ll still be a civilization here to greet them, the way things are going. Incipient fascism, wiretapping, DRM, “trusted” computing … welcome to City 17.
On a minor tangent — SETI is doomed. 50 years from now, all our communications will be compressed, encrypted, Reed-Solomoned, and spread-spectrumed into one big fuzzy-sounding mush.
If SETI’s goal is to decode alien signals then yeah, it may be doomed.
If SETI’s goal is just to detect alien signals, then no amount of compression or encryption is going to be a problem – because the error correction codecs like Reed Solomon are going to have to be applied *last* before transmission, and the output of an error correction codec is always much less random than the input.
Of course, one of the best ways to reduce error rates in a long distance transmission is to focus more transmission power on your target – in which case unless Earth is colinear with the transmitter and target, there would be nothing for us to see. Broadcasting is wasteful.
What we need to understand is where to draw the line,
Simple: wiretapping is illegal. It’s a felony. That’s a law on the books. If government officials need to tap conversations, they can either get a warrant that allows them to do so without being criminally liable, or they can commit a felony and suffer the consequences.
and what kind of oversight and safeguards we need to keep our government near the line we have drawn.
The oversight we need is called the “Rule of Law”. We’ve got laws that cover these crimes; all we need now is a legal system willing to enforce them equally.
Democracy isn’t just about my having one hundred millionth of a say in who runs the country, it’s about having who runs the country be subject to the same laws that I am.
Ed wrote: “The second technology change is the use of computer algorithms to analyze intercepted communications.” A wiretap side benefit resulting from advances in technology may allow us:
1. To use voice recognition (for voice communication) to identify who is actually talking. That also means yet another database of known voices. For printed communication, we would have a library of pet phrases (syntax) that a known writer uses. (This was already been done for a book written by anonymous author, I forget the title.)
2. Measuring signal time delays may allow the wiretapper to pinpoint the locations of those communicating.
“Techology is an ongoing game of leapfrog. Our hope is that our security agencies can leapfrog ahead when presented with a new circumvention.”
I don’t hope that because the good uses that their technology can be put to (catching bad guys) are out weighed by the danger that one day the term ‘bad guys will be defined to mean anyone that does not agree with the current regime. That is fascism, to which we could slip. I don’t think we are their yet, but it would not be an eggageration to say that we are in an incipient stage of the corporate fascist state. (suggest reading of Oryx and Crake)
“Ed has also pointed out that DRM technologies can be embedded in your content stream. Thus a subversive and encryptic message can be hidden from obvious detection. So one of the problems to be faced by the NSA is separating clear communication from hidden secret messages. Encryption does not necessarily have to be strong, just not obvious.”
Yes steganographic file systems, much chaff, little wheat. That is one reason I would propose drowning the net with encrypted files. Create more chafff….
Eric Smith wrote: “The lack of information was not a problem for predicting or investigating the 9/11 acts, but rather the inability to communicate and process the information properly.”
Eric Smith evidently hasn’t heard of PNAC. I suggest you bone up before quiz time (which will be the 2008 federal elections).
enigma_foundry wrote assorted stuff about the NSA factoring large numbers.
Let them factor large numbers. Most major encryption schemes in widespread use right now depend on the discrete logarithm problem instead. 🙂
cm wrote: “The primary point of wiretapping is not to produce court-proof evidence, but leads for further investigation and/or securing “hard†proof.”
Fruit of the poisonous tree is still inadmissible. Don’t you watch Law & Order? 🙂
Ed, I did not mention ‘Enemy Combatants’ or even Guantanamo.
How can you discuss wiretapping unless you first understand why on earth anyone should be concerned about it?
The issue we’re talking about here is thought crime – the idea that the state, from close observation of its citizens, is able to detect and remedy those who veer too close to trajectories considered to present a clear and imminent danger to society and/or the state that protects it.
You start off with bomb-making, contraband images, drugs, IP piracy, and before you know it, you extend it to behaviour exhibiting evident psychosis, inevitably deteriorating into psychopathy if not apprehended immediately.
1) Eliminate terrorists
2) Eliminate those who tend to become terrorists
3) Eliminate those who support terrorists
4) Eliminate those who demonstrate against terrorist counter-measures
5) Eliminate those who demonstrate against the state and its protection of the people
What on earth is wrong with wiretapping otherwise?
Actually the wording was along the lines of “non job related use, especially porn and hate speech”. You figure out the analogy.
Eric Smith: Have you ever considered that the things that you named, and others, are to an extent pretexts? In a previous job, my employer issued an “internet use policy” banning downloads of porn and hate speech. Pressed by startled questions, they admitted in private it was meant to intimidate workers, and in actuality the largest concern was daytrading and checking out stock prices, which really was a big productivity killer, but they felt they couldn’t issue a policy against that. OTOH nobody would seriously break a lance for porn and racism.
Neo (May 8th, 2006 at 1:02 pm): The largest hurdle in investigating anything is figuring out where to look. The primary point of wiretapping is not to produce court-proof evidence, but leads for further investigation and/or securing “hard” proof.
The word “Terrorism,” as used by the people in power in the United States, is much the same as the word “Piracy,” as used by the people in power in the music industry.
These words are part of a visceral marketing campaign. The same sorts of PRopogandists that sold you the “USA PATRIOT Act” (to bypass the checks and balances that patriots stand for), and “Clear Skies Act” (which repeals or reduces air pollution controls), bring us outrageous bogeymen named “Terrorism” and “Piracy,” as well as innocent sounding terms like “Digital Rights Management” and “attrited.”
Even beyond the soundbyte, these words are effective in suppressing and stupefying debate and rational thought. Who needs to *think* about an issue when you have already been told how to *feel* about it?
If you don’t agree with the RIAA, you are a pirate.
If you don’t agree with the political elite, you are a terrorist.
Fortunately, there are great sites like http://www.freedom-to-tinker.com/ 😉 and http://copyfight.corante.com/ to refute the former, and sites like http://scholarsfor911truth.org/ and http://georgewashington.blogspot.com/ to expose the latter.
Thank you, Professor Felten, for standing up for our freedoms.
Techology is an ongoing game of leapfrog. Our hope is that our security agencies can leapfrog ahead when presented with a new circumvention. Ed has also pointed out that DRM technologies can be embedded in your content stream. Thus a subversive and encryptic message can be hidden from obvious detection. So one of the problems to be faced by the NSA is separating clear communication from hidden secret messages. Encryption does not necessarily have to be strong, just not obvious.
My biggest concern about the direction this is going is that: we are constructing an unimaginable powerful infrastructure of control.
This control infrastructure will someday fall into the hands of a malevolent force. When that happens, how exactly will that malevolent force be pried away from these control infrastructure? The route of revolution or ordinary overthrow which had been used in the past may not work.
The only answer I can see is to not build the infrastructure in the first place.
Another factor to be very aware of is: when this infrastructure is built it will enormously tempt those who want to misuse it into gaining political power. Since political power in the past did not include the such a control infrastructure, winning political power now becomes a tremendous prize.
One man, one vote, one time…
“Unfortunately moving to stronger encryption is no solution at all, because then is raises a red flag unless 60% of your fellow countrymen also move to encryption.”
The possibility then becomes to flood the network with strongly encrypted packets, as a form of civil protest. This would not be that difficult to do, I would think if enough people would get on board and start doing it. Perhaps freenet could be a good starting point.
“Plus, if I’m not mistaken, all forms of public key encryption today rely upon to perceived difficulties in factoring large numbers. If NSA has discovered a new factoring technique, what we (the public) thinks is safe could be trivially broken (eg 1970s when NSA knew about differential analysis and nobody else did).”
I suspect they must have made some advances in this regard also. I doubt it is trivial but there may exist ways to use massively parallel processors in ways that would break relatively quickly the smaller size keys. The tell-tale thing for me is that the interest that NSA has for schemes that do not rely on factoring numbers any more but some other kind–elliptical curve cryptography or even n-dimensional lattice vector schemes. But this interest could be just a hedge.
It would be interesting to note what type of work was done by those students in which the NSA has expressed interest ….
There are of course several ways to test whether or not the NSA has actually broken (or can break) large factor based schemes…
I agree with Steve R.: Why would terrorists use our phone system when they can have secure voice communication anywhere there is Internet access?
This is a spy-vs-spy situation, and my concern is that the law enforcement technology will be used for the more obvious uses that invade our privacy while the more difficult but more important national security features will get short(er) thrift. Ditto for your reply, Ed, about collecting more information rather than improving the processing of information.
Steve & Peter,
Unfortunately moving to stronger encryption is no solution at all, because then is raises a red flag unless 60% of your fellow countrymen also move to encryption.
Plus, if I’m not mistaken, all forms of public key encryption today rely upon to perceived difficulties in factoring large numbers. If NSA has discovered a new factoring technique, what we (the public) thinks is safe could be trivially broken (eg 1970s when NSA knew about differential analysis and nobody else did).
I find it exceedingly aggravating that the administration’s response to anything bad (e.g. terrorism or child pornography) is to try to increase the amount of information it collects, despite the fact that the lack of information was not a problem for predicting or investigating the 9/11 acts, but rather the inability to communicate and process the information properly. Collecting more information doesn’t help that at all, and quite likely makes it worse.
Detentions of “enemy combatants†are an important issue, but not one I’m addressing here.
Unless one is deemed an “enemy combatant” for posting a few megabytes of random bits for public download. It seems to me that more and more people are picking up that they should not communicate in the clear — SSL to the mail servers, SSH connections for CVS and Subversion and remote login and tunneling VNC, encrypted disks for data that might get stolen. And because we routinely compress data before encryption, there’s much less redundancy to work with in the input data, and it looks all the more like simple noise.
On a minor tangent — SETI is doomed. 50 years from now, all our communications will be compressed, encrypted, Reed-Solomoned, and spread-spectrumed into one big fuzzy-sounding mush.
As an integrating follow-up to Peter’s post; technological improvements (while aiding the NSA) will also enhance the capability of those who wish to circumvent NSA eavesdropping. Instead of a simple analog or digital data stream, which can be easily analyzed; those wishing to avoid having their conversations monitored could use encrypted data bursts, multiple data streams on different “channels”, and/or time shifting (data out of order). All that we can hope for is that our technology will be “one step ahead” and that those charged with protecting our national security will have a high degree of integrity.
I’d like to respond a little more directly to Prof Felton’s original post here. I think we have to accept that terrorism represents a new and different sort of threat than the traditional criminal element. That being said, I would support a lower standard than “probable cause” to support a wiretap warrant. Especially as we deal with individuals overseas where it could be very difficult to develop probable cause without the wiretaps. I would suggest that perhaps a “reasonable suspicion” standard would be appropriate in this case. I would also suggest that it is perfectly fine for these decisions, taps, and the precise technology being utilized to be kept secret for a period of time (much like in the case of the FISA court).
However, it is extremely important that an understood standard is developed and published. It would also be important that there be rules put in place regarding how long such a tap could continue if no suspicious activity is observed, and that all materials be destroyed for taps where no dangerous activity was uncovered. Additionally, there need to be strong penalties for violation of these provisions.
I think if we could develop a known set of rules for a program like this to operate under, the program could become a valuable intelligence asset for the country and still have some safeguards against abuse. For instance, the possibility of Congressional review is a strong dis-incentive for politicians in the executive to push the limits too far, as the partisan nature of Congress would make such investigations likely in the case of abuses.
Crosbie,
Detentions of “enemy combatants” are an important issue, but not one I’m addressing here. I think it’s important to have a dispassionate discussion of technology and wiretapping, separated from the debate over the Bush Administration’s actions. This is not to say that the Administration’s actions are unimportant. It’s just that others have discussed them at great length, and I’m trying to talk about something different.
To be clear, I’m not saying you should stop talking about those issues. Feel free to talk about anything that you think is relevant. I’m just explaining why I don’t plan to address those issues directly.
Brad,
I agree that if wiretapping gets cheaper and easier, we’ll get more wiretapping, all else being equal. Which is a good argument that all else shouldn’t be equal. You’ll see that theme playing out in my future posts.
What’s more interesting, I think, is that technology changes the kinds of eavesdropping that can happen, in interesting ways. Suppose, for example, that technology allowed more precise “minimization” of wiretap information. (Minimization separates the wheat (real information about crimes) from the chaff (stuff not of legitimate interest to law enforcement or intelligence agencies) in intercepted communications.) This might allow more wheat to be captured, relative to today, without increasing the harm caused by wiretapping. Whether this turns out to be true, and how it affects safeguards and the risk of abuse, are complicated questions which I’ll get around to eventually, after working through some of the easier cases.
Perhaps we should pollute ourselves?
The price of freedom could require mass civil disobedience that involved willfully distributing, circulating, and communicating contraband digital content and intercourse.
Perhaps http://freenetproject.org isn’t quite so crazy after all?
“Have you recently discussed insurrection? A fatwa against the west? Remember, it is your duty as a citizen in a free and open democracy to engage in such discussion.”
If we are all clearly and demonstrably suspects, then suspicion cannot be tolerated as due cause for incarceration without trial.
As noted in some prior comments… criminals, corporations, and maybe even individuals who desire privacy will explore two-way encryption. Will using encryption for voice communication remain legal? Or else how else will government counter this?
This form doesn’t seem to like my formatting, here is the rest of my above comment;
However, there is somehwat of a consesus that it would require less than 1 terabyte of storage to store every word spoken, from birth to death, of the typical single human today – of average life expentency.
So, the technology to record everything from this point forward is here today, and for cheap. This begs the question, how long should this data be retained? If the data can be retained indefinitely, then the data can be re-sifted every so often as analysis methodolody advances.
We may see the same syndrome as happened with DNA technology, when retained evidence from years or decades ago could be analyzed with new technology. Perhaps we all should just think very carefully before we say anything.
The issue of ‘archival’ raised by Jeff is an important one. There is considerable debate about how much storage would be required to store every word spoken by every human in all of history. The guesstimates range from 5 exabytes to 400 zettabytes.
However, there is somewhat of a consensus that it would require
Collecting and analysising private communications is not the worrying thing, it’s that people can be arrested on vacuous charges and incarcerated without trial.
What a wonderful world it will be when computers print out arrest warrants for suspected terrorists/paedophiles/copyright infringers, based merely upon heuristics and ‘expert’ systems.
Lock ’em up first for safety, then one day figure out if there’s actually a case against them… if budget permits. ‘No smoke without fire’ ensures popular support.
The best defense against the state collecting citizens’ private communications and abusing it, is for everyone to communicate publicly, e.g. Google to provide an open version of gmail (all mail open to inspection and data mining by anyone).
The technological change would be interesting to present to the drafters of the constitution and their early electronic age successors.
They wrote protection from searches into the 4th amendment, and later protection from electronic surveillance was fairly deemed as implied by that. All of this was written when the state’s surveillance powers were tiny, and surveillance was very expensive, and most importantly, didn’t scale.
Yet even then, the need from strong protections from it was clear in their minds.
Now that surveillance is so much easier, and can scale to a level where everybody can be watched if we allow it, doesn’t that suggest we would want to write in _more_ protections and not less, even if all we are trying to do is retain some of the balance?
It isn’t going to be a one-sided arms race with only the NSA benefiting from Moore’s Law. On the flip side, the amount of crud for them to intercept and analyze will be increasing as fast as their capabilities to intercept and analyze stuff, and on top of that, the truly dangerous crooks will have increasing access to methods of communicating securely. At best, the NSA will catch the dumb crooks (easily caught anyway) while driving the smart, dangerous ones to wholesale use of PGP-encrypted email…not to mention, the dumb crooks can’t be prosecuted with the illegally obtained wiretap data anyway.
Let’s assume that the NSA will be technically capable of recording and analysing all (domestic + international) phonecalls in ten to twenty years. Should they be allowed to do so? Will it help security? How big is the threat of such a spy program to freedom and democracy?
As an engineer I know that voice recognition and natural language processing are hard problems. It will be fair to assume that there’s a 1% error rate in the automated processing… How many false alarms will it generate compared to the amount of terrorist acts detected? Can smart criminals evade detection by using some form of scrambling or encryption?
The current trend of “anti-terrorist surveillance” is a danger to democracy as it becomes more and more dangerous to talk about sensitive topics. The STASI (former East German Secret Service) collected typeproofs from all typewriters sold and kept a register of whom bought which. The US government coerces computer printer makers to add identifying dot patterns to all printed output. Draw your own conclusions…
For me, a very real concern is that this information (possibly the *actual* voice conversation) will be archived in some electronic store. As the capability to store and translate (vocie to text) get better, there may be a tendency to store and record more information (that is, calls) then before. As we have seen recently (pick your favorite lost backup tape or unauthorized database access), this information whether by accident or design may be entered into the public domain.
In order to protect the privacy of those being tapped, rules against storing data beyond it’s useful life must be in place, as well as, controls on access to the data (perhaps a beneficial use of DRM?) must exist.
. . . question regarding the application of Moore’s Law. I understand why Moore’s Law dictates that storage capacity will increase. I also understand why Moore’s Law suggests that the government will be able to apply its algorithms to a larger quantity of communications.
I am less clear on the relationship between what Moore’s Law says regarding computing capability and the perfectability of the algorithms being used to identify specific communications for human review. In other words, in the “new world of wiretapping” we are hypothesizing, is it our premise these algorithms actually work?
I ask because there are probably a fair number of people whose view of what the extent of national security wiretapping should be would be substantially swayed by whether the NSA’s datamining actually works. I also ask because, as the Washington Post reported, the FBI’s major gripe with the NSA’s program is that it generates large quantities of fruitless tips.
Good points.
I share Jeff’s view that the NSA itself won’t disclose what technology it is using. It took the New York Times to reveal that the program even existed, and the Bush administration has been incredibly resistant to revealing even the most basic information about the NSA’s program.
(It is this resistance that makes me especially curious what technologists can discern about the NSA’s current capabilities. I admit this is tangential to Professor Felten’s main objective, but I find it so interesting that I made what I now recognize was an inappropriate attempt to redirect your collective brain power to this question.)
So, to stick to the topic, I do have one
Regarding Itinerary Lawyer’s question about what the NSA can do today, I agree with Jeff. What matters most is that thanks to Moore’s Law, the NSA will eventually be able to record and analyze all phone calls. We should be thinking now about the implications of that fact.
But you’re still asking, I’m sure, what the NSA can do today. We don’t know for sure. If I had to guess — and remember this is no more than a guess — I would say that they can’t record and analyze all cross-border phone calls today, but that they have to limit themselves to a subset of calls, selected by factors including the locations and phone numbers of the parties to a call.
I think the questions posted by Itinerant Lawyer really aren’t going to be answered. To think that the NSA is goign to let the public in on the technology they are using for this sort of analysis is foolish. However, I think Prof. Felten’s point about Moore’s law is this: whatever the state of their technology now, it’s going to get better and very quickly. So, the only proper ways to set limits on this sort of tapping that won’t be irrelevant before their written is to assume that these agencies can store as many conversations as they like and that they can use text-based analysis on all of them with 100% accuracy. To assume a technological barrier is a good protection of rights is foolish, especially given how quickly technology changes.
That being said, I think the other posters have both mentioned judicial oversight. This is a crucial piece of protecting the rights of citizens and setting proper boundaries here. The question is, what standard is being used to determine when an tap order is appropriate, who has the authority to make that decision, and what safeguards are in place to ensure that no one is performing these sorts of taps without the appropriate judicial review.
I am glad you are blogging on this hard question. It would be helpful if you were to begin by discussing a threshhold question, specifically what the NSA is technologically capable of doing.
Here are some interesting questions:
(1) Can the NSA apply the same techniques to analyze voice communication that it can use to analyze email? This question probably boils down to whether the NSA now has speech recognition software sufficiently advanced to convert spoken words (in many languages) to text in some efficient manner. If it is not sufficiently advanced, what else could the NSA be doing in addition to analyzing time/frequency/destination of calls?
(2) How accurate do we think these algorithms are?
One salient issue here is storage. Computers have become de facto surveillance devices that record every correspondence and interaction online. They mock the Fifth Amendement right to not incriminate oneself.
Another issue is legal oversight. Surveillance will always exist. We must assure that it is employed carefully and within the bounds of the Constitution rather than wholesale, in order to compensate for lack of human intelligence sources on the ground.
There are a few yardsticks I’ld like to put in the sand to direct the discussion: In a democratic country, all branches of government are populated by “ordinary people” that are asked to do their job in governing the country for the people. Government officials should justify their actions to “the people”; in a democratic country an independent press plays an important role in disseminating (dubious) government actions and justifications. The problem with the “security agencies” and their actions is that these agencies claim that they are allowed to act in secrecy and so avoid justifying their actions.
I can see that phone taps are a source of intelligence, both for security agencies, competitors and ex-lovers. So, with every phone tap the government has to balance the interests of the general public against the interests of the persons tapped. In cases where there is a concrete suspicion of criminal acts, the balance will easily fall to the side of setting the tap; I see this balance fall to the side of not doing “wholesale” tapping of trunks of connections.
Should the NSA, CIA and FBI be allowed to set taps, yes, in individual cases. Should there be legal oversight, yes, off course. Should these agencies disclose statistics and procedures, yes, to allow a public discussion.