Schneier on Security
A blog covering security and security technology.
« Evercookies |
| Friday Squid Blogging: "Truck Carrying Squid Crashes In Broccoli Field" »
September 24, 2010
Real-Time NSA Eavesdropping
In an article about Robert Woodward's new book, Obama's Wars, this is listed as one of the book's "disclosures":
A new capability developed by the National Security Agency has dramatically increased the speed at which intercepted communications can be turned around into useful information for intelligence analysts and covert operators. "They talk, we listen. They move, we observe. Given the opportunity, we react operationally," then-Director of National Intelligence Mike McConnell explained to Obama at a briefing two days after he was elected president.
Eavesdropping is easy. Getting actual intelligence to the hands of people is hard. It sounds as if the NSA has advanced capabilities to automatically sift through massive amounts of electronic communications and find the few bits worth relaying to intelligence officers.
Posted on September 24, 2010 at 1:23 PM
• 27 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
It sounds as if the NSA is claiming that they have the capability to automatically sift through massive amounts of electronic communications and find the few bits worth relaying to intelligence officers.
It sounds as if the NSA is refusing to be said to be claiming that they have the capability to automatically sift through massive amounts of electronic communications and find the few bits worth relaying to intelligence officers.
It sounds like the NSA wants the president to believe they have the capability to automatically sift through massive amounts of electronic communications and find the few bits worth relaying to intelligence officers.
I'm not buying it.
If this was really the case, wouldn't we see more of the "bad guys" being caught?
Identify a "bad guy", get lots of troops in armour, surround his house/apartment ... and listen into his calls when he panics.
Have more troops ready to move on those calls with the same procedure.
Sorry Bruce, but I don't agree with you on this one. Assuming the NSA has developed a new advanced ("real time") capability based solely on fourteen words and three sentences of a quote from from a briefing is a leap of faith. I would hardly call that proof.
That more "bad guys" aren't being caught is probably a reflection of the fact that there really aren't that many of them.
As far as the NSA's statements, if they think that claiming some bullshit might deter some guy living in a cave in Afghanistan or the Sudan from making a cell phone call, they'd do it in a heartbeat.
Bob Woodward isn't much of a filter either. Even if he knows its bs, he'd gladly pass it on in his quest for the next piece of Washington insider gossip.
Or : sounds more like they can get intercepts of a specific person or group of people, process them and get results to agents in real-time. Like in the Bourne Conspiracy.
As opposed to "over the last month, he/she/they said X, Y and Z."
Do you use Tor?
Review the Tor nodes list, updated daily:
Search within for:
nodes, now imagine how many nodes are fronts for a corp or gov/mil entity hosted in different countries under different names
now consider the policies of data retention in many countries
consider data "sharing" between countries
how many Tor nodes aren't run by intelligence ops or front groups for them?
Do you mean 'apart from those exit nodes run by criminals harvesting credentials'?
Good point about the difference between eavesdropping and getting the relevant data to analysts Bruce. The sifting process itself is problematic I imagine--I'm thinking of the ways the sifting occurs involving more than one fallible human. I wonder what the audit process is for the sifting part?
The sifting process surely leaves lots of room for error or in domestic contractor run (ATT Security) terror watch operations lots of room for waste, fraud, and abuse of DOJ funds.
"It sounds as if..." doesn't mean that Bruce believes it. "It sounds as if..." might mean "they make it sound like..." which could be spin for the benefit of the potential POTUS, political rhetoric for the benefit of the Director's career, whatever. The post on this blog doesn't claim this piece of information as definitive.
I'm certain that the high-level people at NSA believe they have this awesome, infallible capability to listen, analyze and react perfectly to what they intercept. I personally believe there's no possible way they could correctly analyze and properly react in anywhere near real-time.
I suspect that the reason more bad guys aren't apprehended is that this is a pointless endeavor. Codewords and phrases would render moot this kind of automated task, not to mention most people aren't 'bad guys'.
I *fear* that the reason that such apprehensions aren't publicized (assuming that this program *does* function, correctly or not) is that they want to keep the secret. Because it's unconstitutional, because they make mistakes, etc. You catch a bad guy with this and you're wrong, it's a media and political nightmare *and* people know that you're running this program.
Too much 'raw intelligence' can easily delay getting real information into the right hands. Still, I wonder when the govt. will turn this new technology on American citizens when they exercise their freedom of speech.
Two words: Lattice Theory.
All of these responses assume a Top-down approach similar to parsing. Such an approach requires that you know what you're looking for and to run a prohibitive amount of queries against an enormous database.
Instead, by building a lattice of concepts based on relationships between intercepts, and by intersecting with the appropriate ontology, only a minimal number of results should have to be considered.
Schneier, it seems is lacing his blog with mere propaganda on behalf of the U.S Intelligence agencies.
Keep up the good work,
I don't know how it is so difficult to mask your true meaning during communications or how we can get any intelligence. All the terrorists have to do is establish a simple code. Disguise yourself as a purveyor of produce. Substitute "lettuce" for "nuclear weapon" then talk about the delivery arriving at the World Trade Center in New York City "by Fed Ex" at 9 am next Friday and I doubt many automated systems will realize what is really being said. It gets easier if I am the suspect and they know I don't sell lettuce, but if I am just another message sender on the internet who is being monitored for use of key words, no one will figure it out. This adds to the problem, the need to actually analyze messages and conjecture about meaning.
for all the billions spent on so called intelligence, we based a war on curveball's information without our intel people ever being able to talk to this shill for chalabi or israel or both. Intelligence agencies are political think tanks so of course they are scamming against the people of our nation, If another 911 is being planned, there is little chance it can be prevented by these torturers and fools.
The same article says that the cia has a 3000 man highly trained antiterrorist unit operating in packistan and afganistan, you must take this with a shaker of salt, considering that it comes from the guys who paid afgan "allies" to "capture" binladen at tora bora, and who ran spectacularily double crossed spying operations in cuba and east germany for years not to mention their "success" in spying on albania through out the cold war. They are accomplished at spinning their failures as successes.
Unfortunately, I think the NSA has made progress along these lines and will make more progress in future. That's bad for all of us. You may think that Obama will just use this capability against our enemies - and you may be right - but the capabilities will not go away when another Nixon, or worse, gets into the White House. These capabilities will be used, sooner or later, to snoop on loyal American citizens who happen not to agree with the person occupying the White House at the time.
Given that these capabilities, if they really represent any serious advancement, were spoken of almost two years ago... I'm certain that they've been used against American citizens already.
Person to person is so much safer. Don't discuss private things online, on phones, etc.
From a post about something I am working towards:
"The need for FOSS intelligence tools for sensemaking etc."
"As I see it, there is a race going on. The race is between two trends. On the one hand, the internet can be used to profile and round up dissenters to the scarcity-based economic status quo (thus legitimate worries about privacy and something like TIA). On the other hand, the internet can be used to change the status quo in various ways (better designs, better science, stronger social networks advocating for things like a basic income, all supported by better structured arguments like with the Genoa II approach) to the point where there is abundance for all and rounding up dissenters to mainstream economics is a non-issue because material abundance is everywhere. So, as Bucky Fuller said, whether is will be Utopia or Oblivion will be a touch-and-go relay race to the very end. While I can't guarantee success at the second option of using the internet for abundance for all, I can guarantee that if we do nothing, the first option of using the internet to round up dissenters (or really, anybody who is different, like was done using IBM computers in WWII Germany) will probably prevail. So, I feel the global public really needs access to these sorts of sensemaking tools in an open source way, and the way to use them is not so much to "fight back" as to "transform and/or transcend the system". As Bucky Fuller said, you never change thing by fighting the old paradigm directly; you change things by inventing a new way that makes the old paradigm obsolete. ... As with that notion of "mutual security", the US intelligence community needs to look beyond seeing an intelligence tool as just something proprietary that gives a "friendly" analyst some advantage over an "unfriendly" analyst. Instead, the intelligence community could begin to see the potential for a free and open source intelligence tool as a way to promote "friendship" across the planet by dispelling some of the gloom of "want and ignorance" (see the scene in "A Christmas Carol" with Scrooge and a Christmas Spirit) that we still have all too much of around the planet. So, beyond supporting legitimate US intelligence needs (useful with their own closed sources of data), supporting a free and open source intelligence tool (and related open datasets) could become a strategic part of US (or other nation's) "diplomacy" and constructive outreach. "
The NSA is dead serious about this, and began work at least as early as the '80s on it, when I stopped working with them professionally. Presumably, even fairly lousy efforts would have progressed a lot since. Backing out the spin, I'm sure that they still have something at least moderately credible along these lines, and yes, of course, are using it. And "you can't bust me for breaking the laws, what I do is a national security secret" is and has been going on for a long time now.
I also know how hard it is, having worked with it, and the crop of ridiculous, ineffective bad guys our people have "caught and prevented" shows how few really dangerous ones there are. I mean, these movie plot (at best) preventions they have done make me laugh -- and even with that, no real serious damage (other than to our liberties) has been done by the real bad guys in quite awhile, either.
Waste of time and money that puts (again) the wrong men in charge of other men, using tech and nature as the tool. As CS Lewis said -- man's conquest of nature is a spin on man's conquest of man, with nature as the tool....
Just like things such as the drug laws, whatever else you think of them, make the very worst people rich and powerful -- and they become supporters of those same laws. Unintended consequences, but once in place, real hard to change with that sort of backing.
"I'm not buying it.
If this was really the case, wouldn't we see more of the "bad guys" being caught?"
What? More bad guys are being reported to us daily due to this technology.
How long does it take for a reasonably intelligent reader to figure out if the comments at this blog should be taken seriously? I wouldn't underestimate the value of common sense and experience at the NSA. They've got the same Internet as we do, plus some powerful surfing tools. Is it so hard to imagine that they can read critically?
I don't believe it. If such a thing were operational today, then it must have been running for a few years already given then size and scope of a system covering the US IC apparatus. That being the case, why haven't we 'won' yet, if this system is so great? The fact that the political hand wringing about domestic security continues unabated leads me to think this is nothing more than some bullets read from some contractor chart-o-graphic about how wonderful they think they will be after spending X billions of our dollars trying to emulate Minority Report. After all, it's election season.
Anyone who believes this has failed to understand either the magnitude or the inherent difficulty of the problem.
The magnitude of the problem is astonishing: just to monitor all domestic communications within the US alone, there's absolutely no way the federal budget would stretch to cover it (and I mean the _entire_ budget, if the government didn't have to spend a dime on anything else). To do it with human listeners, you'd have to employ something like sixty million people full time, maybe more.
As for difficulty, it'll knock your socks off. First off, computers are nowhere near smart enough to do useful filtering of the type discussed, and anyone who tries to tell you otherwise has never tried to program serious AI. The problem is *hard*. Frankly, it's harder to get right than natural language translation, which is infamous. Sure, with some clever algorithms you might manage to get it right 45% of the time, and you might get lucky and be close another 25% of the time, but the computer won't have any idea *which* 70% of the time it's right or close to right, so a human is still going to have to review 100% of the results. The real kicker, though, is that most humans can't do it consistently either. After a few minutes the eyes gloss over and the brain starts thinking about something else.
If the described problem is even theoretically possible to solve, we need at least another fifty years, minimum, of technological advancement before we can *possibly* get serious about tackling it. Several hundred years is more likely. Never is also a possibility, as far as we know at this point.
> It gets easier if I am the suspect
That doesn't make the problem easier: it bypasses the problem entirely. If they already know whom they need to watch, there's no need for a vast listening network or any kind of sifting at all: they can just bug the suspect's phone (and house, and whatever) and have the officers who are working the case listen in on the suspect directly. No problem.
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.