Hacker News new | past | comments | ask | show | jobs | submit login
CIA ‘Siren Servers’ can predict social uprisings 3-5 days in advance (sociable.co)
160 points by andrewke on Oct 8, 2016 | hide | past | web | favorite | 91 comments

I was personally in Western Tunisia (on the Algerian border) as the Tunisian uprising, the first of the Arab Spring, began. I was there a month. The protests were in full swing for weeks in every village in the west before anything hit western media. No travel warnings or any of that backdated pretend bureaucracy. I very much doubt the CIA had much of an idea what was going on outside the capital. The only western presence I saw during the whole period was a German diplomat taking his friends to see some ruins further east, which I doubt he would have been doing if he knew there was serious unrest. Internet is barely utilized there - how are they going to gather intelligence? It really gives you food for thought. Stay off the wires (including finance), and it's very hard for them to see much of anything.

And the other way round. I lived in Beijing at the time when Americans bombed the Chinese embassy in Belgrade (May 1999). CNN and other Western news channels were showing films of furious Chinese protesters shouting and throwing things at the U.S. embassy buildings in Beijing. This went on for days.

When I actually went to the Sanlitun embassy area, the closest thing to a riot was a stout Western woman walking her poodle outside the German embassy. Otherwise, it was too hot for anyone to bother go outside and everything was sleepy. The week's worth of rioting on CNN was shot during one day in a well-organized riot properly supervised by the local security officials.

There is no doubt the Chinese were genuinely offended, officials and public alike, but they wouldn't let such a thing cause disorder even though CNN etc. needed the drama.

The Chinese were far from being genuinely offended, they were more than happy with the parts of the spontaneously disassembled F-117A. Probably they even helped to make the new artificial lake afterwards.

What on earth are you talking about?

The Chinese got a complete crashed F117 during the NATO bombing.

So are you saying that the embassy was bombed to stop the transfer of F117 parts? Because that will need a source

There were various related rumors circulating at the time. You can draw your own conclusions but please allow others to share what is only available as oral history.

My personal favorite (heard in Beijing in the early 2000s) was the story that the Chinese military were monitoring the newly deployed American stealth fighters during the Belgrade bombing and had managed to get radar lock on one (or more) of the F-117As from their embassy. The Americans, the story goes, wanted no more of this and bombed the Chinese embassy.

According to The Font of all Knowledge [0] three Chinese "reporters" were killed in what the CIA admitted was its own bombing operation. It has long been common knowledge that the Chinese government uses the Xinhua news agency as a cover for clandestine operatives [1].

True or false, it'still an inventive rumor.

[0] (https://en.m.wikipedia.org/wiki/United_States_bombing_of_the...) [1] (http://www.nytimes.com/2000/02/20/world/in-watching-hong-kon...)

A more effective strategy to constant surveillance is to simply raise the background level to just below actionable threshold, so that way their intelligence is effectively useless. That is, really, you should just sit around all day and talk about how you dislike what they're doing, think they're career criminals betraying the nation, that you want to actively do something about it, etc. Bonus points for mixing that conversation with discussion about how you'd like to tour DC and see the various monuments, including the Washington monument, Lincoln memorial, and WWII memorial.

If the background unhappy chatter is close to actionable (or is in the mind of the classifying algorithms), then actionable chatter (eg, actually planning an attack) can only be distinguished from background chatter near when the plans become "active" in some sense, and likely too late to meaningfully respond to. (To use a physics analogy: if you try to predict cavitation in flowing water really close to boiling, it's harder.)

In many senses, simply embracing openly that we're very angry as a society, even in the face of constant scrutiny, prevents that same constant scrutiny from determining when we're planning something. It can only really tell that we're pissed off, in general.

Maybe you could create AI agents to flood the Internet with seditious chatter.

Amongst many, IBM has done something that does this and the opposite. It was funded by public money (DoD).

That sounds pretty interesting. Where could I learn more on this?

And then start an actual riot by mistake?

I mean, the whole point of that strategy is to provide statistical cover to prevent the detection of forming riots (or attacks) ahead of their actual nucleation and rioting (or attack). Effectively, you give the appearance of spontaneous formation of riots out of a near-rioting background, even though the actual formation still takes coordination ahead of time, by simply having a lot of times you don't actually plan a riot, you just say all the words to (preferably to random people).

So I wouldn't call it 'by mistake', more like 'by design'.

Of course a social uprising in Tunisia is very unlikely to spread to the USA. Movements in the UK, Australia and Canada however...

As for staying off the wires. Thanks to the network effect you need to have a large portion of your network do the same.

How is a social uprising in Tunisia more likely to spread to the UK or Canada than the USA?

They're saying that Tunisian movements are unlikely to spread to the US, but movements in the UK/Canada can spread to the US.

Thanks, that makes more sense.

The protesters didn't have cell phones? Just correlating their locations would tell you something was up.

I would say CIA likes to organize those, so maybe they can use those pretend 'Siren Servers' to claim they can sense unrest.

Probably they can detect uprising in Germany...

It's interesting given that many citizens have predicted that things are about to collapse.

The difference was in the number of police man in the coastal cjties. They were non existant which suggests that the regime was fighting in the unrest locations at full capacity. It never happened before.

> Stay off the wires (including finance), and it's very hard for them to see much of anything.

NGOs. [yet again with silent (dumb?) downvotes in the same comment chain that mentions "network effects".]

I don't think I've ever seen a one-word comment make it on HN...

Maybe elaborate on what you mean? The single acronym in your reply doesn't actually expound a point of view.

NGO is very commonly understood to mean Non-Govermental Organizations.

OP proposes that staying "off wires" is a remedy. Simple point made is that intelligence agencies have recourse to wide ranging sensory networks, including NGOs (whether complicit or simply surveilled.)

Except that pointing out that an agency can, with investment, have limited, targeted active surveillance isn't really saying anything about what's effective from stopping that same agency from wide-scale, passive surveillance.

NGOs (even in the case where you're having their help by carrying scanners around for you) aren't within orders of magnitude of the information collected by the average person carrying a cell-phone and tapping the network. You would have to install expensive and complex monitoring equipment over the entire city and install the backend to process it to get even close -- a capability that the US government doesn't have to do everywhere. Which is why they made use of already deployed networks in the first place.

Stopping their passive surveillance strips them of 90% of their power, and NGOs (or other organizations) don't put even a tiny dent in that change.

Which is why many of us were looking for a little more argument than what was given: it's not clear, at all, why either you or the original person thought that NGOs have anywhere near the penetration in to everyday life or surveillance capability that cellphones have.

NGOs aren't used by intelligence agencies as a means of boosting signal intelligence, they're used as a means of boosting human intelligence.

NGOs tend to have overlapping members and social circles as grassroots organizations. They don't need to gather cell phone signal intelligence, they can just ask their staffers "what do you think of these protests?"

Different threat vector entirely.

> Different threat vector entirely.

My point was exactly that it's a different threat to have them have to collect human intelligence from NGOs rather than have every text message, phone call, email, etc from and up-to-the-minute location data on every person they're looking for intel on.

I don't care if the CIA can call up every NGO on the globe and talk to 50 different people there to get 50,000 reports on the current civil unrest across the globe, collate them down in to usable intelligence, and present that to our leaders. That's literally their job. They should do that. Lots of that. All of it they can. (I would be shocked if the CIA didn't already heavily extract intelligence from NGOs.)

It's just a completely different scale from being about to know where every person is minute-by-minute, and everything they say, text, email, read online, etc. forever, to be looked up and cross-referenced whenever you need. Particularly because the conversations with NGOs have to focus on targets of interest, while they can literally just watch every phone and digital transaction. All of them. All the time.

So I'll be blunt: you guys are just wrong that them being able to get any kind of intelligence from NGOs, be it sigint or humint, that would make up for the firehose of digital exchanges being shut down or rendered useless.

Every time I read about expensive programs like these I think of what else that money could have been spent on that would prevent the need for populations to rise up against their governments.

It's the CIA, there is a high chance that they are spending more money to fuel the revolution.

Well, that's one way of knowing when uprisings will happen. Deming wrote that if you have a policy of rewarding people for fighting fires, you also need to watch who is setting those fires.

The cobra effect is a great example of this: https://en.wikipedia.org/wiki/Cobra_effect

You can also get some good stats by lighting some random fires yourself and record who comes in to either put them out, or fan the flames.

The State Department especially are training foreign activists the CIA needs to look no further then there for 80% of where the unrest is fostered from.

gotta have a way to measure your ROI.

Uprisings are nothing but a reaction to an unhealthy situation. Terrorism boils down to that same problem.

Yet, we never seem to be too interested in understanding and resolving the issues we have created which led to these reactions.

Which do you think guarantees an intelligence agency more funding?

My feeling is that there is some conflict of interest here when it comes to fixing societies' woes and people who manufacture tear gas and drones. A lot of money to be made in keeping the status quo, what do you get from fixing the world?

Nothing new here :


But go and download that new social app, citizen. It has new stickers.

> what do you get from fixing the world?

An abundance of profound happiness which, for now, unfortunately remains too invisible for too many people. But I have a feeling that this is about to begin to change now.

So who gets that money? You support democracy in China and the Communists start an uprising. You support Communists and the other side starts an uprising; someone is always pissed off -- it isn't "government" vs. "the people." It's vastly more complex. Rarely are there two sides -- more like 200. Intelligence is a wilderness of mirrors as James Angleton had said.

You could say "end the CIA" but that's sticking one's head in the sand -- as long as there are people, there are going to be bad people and groups. Obviously the determination of what is 'bad' isn't so simple.

This is a complex issue and redistributing money doesn't solve it.

> This is a complex issue and redistributing money doesn't solve it.

Strangely enough, this seems to be the point of view of all the people who have the money. Funny how that works.

Claiming everything is morally ambiguous and thus we need the CIA, which has real examples of unambiguous harm, seems more like sticking one's head in the sand to me.

This is not a moral issue, ambiguous or not.

CIA and its counterparts elsewhere will be here as long as there are nations. Event if every countries were on friendly terms, which won't happen, there would still be conflicts of interest, and hence the desire to gain the upper hand via intelligence.

Dowwie didn't say anything about "redistributing money".

> as long as there are people, there are going to be bad people and groups

Proper education reduces the numbers of "bad" people.

Not to put your statement down, but this probably cost a lot less than that.

“much harder to convey confidence for the policymaker who may make an important decision from advanced analytics with deep learning algorithms.”

This is an unsolvable problem. From the policymaker's point of view saying "deep learning told me so" is exactly the same as saying "god told me so"

"god told me so" worked very well for thousands of years.

It certainly worked, whether or not it worked "well" is up for debate.

defacto, no debate...

"In certain sectors such as finance, it can be self-detrimental as in the case of the financial meltdown in 2007-2008. When a predictive system becomes too efficient, it undermines the very foundation upon which it stands.

If you’ve gathered enough wealth and your services are so efficient that their values skyrocket, but there’s no money left for anyone to afford them, then you’ve fallen victim to the seductive power of siren servers."

what the fuck is this author talking about? this is so incoherent. CIA can predict black swan events now and the super computers can accumulate so much wealth that no one can afford it? what is this website? is this satire?

I can not predict social unrest 3-5 days before it happens but my google alert for "black man shot by police" can do the same reliably 24 hours before social unrest starts.

Suck this CIA.

I think you'll find a lot of false positives, as unfortunately that alert will be sent more than once a week: http://www.killedbypolice.net/ (M/B means Male/Black)

Set an alert for "cop convicted of unlawful killing of suspect"


>Tensing pulled over DuBose near campus for a missing license plate. His attorney said Tensing feared being dragged under the car as DuBose tried to drive away. The shooting was captured on video by Tensing’s body camera and depicts DuBose repeatedly being asked for his driver’s license. After DuBose refuses to produce it and get out of the car, a gunshot is heard.

>Tensing, who was fired from the department, was released from jail after posting 10% of $1-million bond. The county prosecutor called the shooting a “senseless act.” Tensing is scheduled to face a jury starting Oct. 24.

>Los Angeles Police Chief Charlie Beck in January recommended criminal charges be brought against Officer Clifford Proctor, who fatally shot an unarmed black man in the back.

>Proctor told investigators he saw Glenn’s hand on his partner’s holster. But an LAPD report that drew upon security footage from a nearby bar in Venice found Glenn’s hand was nowhere near the holster.

>Michael Slager faces up to 30 years in prison if convicted of murder in the shooting death of Scott, who was running away on foot from a traffic stop in North Charleston, S.C., in April 2015. Scott was unarmed.

>The shooting was captured on cellphone video by a passerby. Slager fired eight times at Scott before he was killed, reigniting the debate over how blacks are treated by law enforcement officers.

>Slager, 34, was fired by the North Charleston Police Department and stands trial in October. Previously held in solitary confinement, he has been free since January, when he was released on $500,000 bail and put under house arrest. Last fall, North Charleston approved a $6.5-million civil settlement with Scott’s family.

>Peter Liang, a rookie New York City police officer, was convicted Feb. 11 of manslaughter in the death of Gurley.

>Liang was patrolling a public housing high-rise with his gun drawn in 2014 when he fired and a bullet ricocheted off a wall, hitting Gurley. Liang said he had been holding his weapon safely when a sound jarred him and he accidentally fired.

>In April, a judge reduced the conviction to negligent homicide and sentenced Liang to five years’ probation and 800 hours of community service.

>Chicago Police Officer Jason Van Dyke is charged with first-degree murder in the 2014 shooting death of black teenager McDonald. Van Dyke was charged in November on the same day that the city, on the orders of a judge, released the explosive dash-cam video showing McDonald being shot 16 times.

>On the night of the shooting, Van Dyke and his partner trailed McDonald for nearly half a mile, from a trucking yard where he was said to have been breaking into vehicles. The officers radioed for a Taser in order to apprehend McDonald, who they said was carrying a knife in the middle of the road. According to the video, the officers arrived 10 minutes after the first call and, within 21 seconds, Van Dyke had emptied his 16-round handgun.


Yes, and what percentage of those killings are actually justified? The police have an inherently dangerous profession.

Justified or not, I still consider those deaths unfortunate.

All lives matter.

And I can predict 3-5 days in advance how many people will come to the protest event by having access to the Facebook list of "who will come" for that event. The people report themselves.

That's more indicative of the agitator movement "Black Live's Matter" and the disgusting mainstream media that assists them in provoking the general populace. The death of black men at the hands of the police is unrelated.

I downvoted because this comment is incredibly inflammatory and instead of providing any evidence relies solely on inflammatory language and emotional appeal.

I'm curious, who are your favorite journalists or what are your favorite news sites?


"We have, in some instances, been able to improve our forecast to the point of being able to anticipate the development of social unrest and societal instability some I think as near as three to five days out"

with that in mind, i would say that they still have a lot of false positives...

Would you trust the CIA's predictions, and that they were accurate rather than self-serving?

Welcome to Psychohistory. Thanks Hari.

No problem, it's just a 30,000 year dark age.

We come out OK at the end of it!

My thoughts exactly. This is no Minority Report. Thanks Isaac!

Likely because they're detecting their own signals.

It's pretty clear by now that for instance the "color revolutions" were instigated in a large part by the US government; social media manipulation was a prominent attack vector.

I don't see anything too alarming here. Predicting social uprisings 3-5 days in advance is easy, as long as you're happy with a few hundred false positives for every solid hit. Signal-boosting is only useful to the extent that it doesn't equally boost noise.

Besides, I'm quite sure that this other initiative is producing far more actionable data, even if it's hardly breaking news: https://www.youtube.com/watch?v=juQcZO_WnsI (2011)

In my limited experience, I feel like the CIA has a lot of room for improvement. Reminds me of this-


There's the problem of guys like Michael Morrell, who have no operational experience but go on Charlie Rose and advocate murdering Russian and Syrian officials. Not that I disagree but why publicly talk about it? Morrell's an economist. The agency being run by political bureaucrats I suppose is inevitable but it's unfortunate.

However, my experience on the science and technology side is also lacking. There's some super smart people who seem to lack any directive of strategy. I sometimes think, is this guy really this ridiculous or is he levelling me?

I love the CIA and am biased in their favor, but I can't think of any personal experience that makes me have confidence in them. Perhaps they are good at operational missions, but they are run by bureaucrats and the science side has the lack of socially smart, well rounded individuals.

The mystique and aura of powerfulness is seductive but the reality of the world is, everyone is much stupider than you can imagine. It's a wonder the world works at all.

>"run by political bureaucrats I suppose is inevitable but it's unfortunate."

So if it is unfortunate, what classification of people, in your opinion, should be running it?

What a weird comment you have posted... You love the CIA? are you a sadist?

I am serious about that question, as while I agree that all states need their intelligence services - the CIA has done some downright EVIL shit. And if you "love the CIA" and are unfamiliar with their history, then you're a fucking moron.

Now, its not to say that the US needs to, in their own best interest, remain as a world super-power... and we shall -- but its to focus on the fact that we can do so in a better fashion.

I take the words "I love the CIA" to mean that you're either a sadist, a child of an op, or young and stupid, and by stupid - I just mean not critically thinking about what you have said/what your understanding of the CIA has done/is.

Again, The US will be the predominant power in the world for the foreseeable future, but have some humanity -- something the actions of the CIA fucking lack.

And if you love the CIA so much, you'll realize that they own this country, and that is not necessarily a good thing, you sound naive.

It's a cultural thing I believe. Some cultures have a different way of dealing with conflicts, and each one of those has its pros and cons.

The US for example threatens adversaries before it acts. And sometimes it will not act even though she says so. (bluffing like in a poker game) The benefits as I see them is being somewhat more transparent.

The Russians on the other hand will not threaten. They want to keep all their options hidden. But when they come out saying they will do something they very likely will and sometimes this can look like an extreme escalation coming out of nowhere.

He has some interesting points. I can't say I agree with his opinion of where the "real value" of a service comes from. No doubt a service would be nothing without it's users, but it would also be nothing without the blood sweat and tears of the developers who built it.

In the end we're creating something that people want to use and giving people something like Instagram for free is all the reward they deserve. If you Instagram where to pay their users they'd go bankrupt. It's not a viable business plan.

I also feel like he's missing the whole point of open source. No one ever said it would distribute wealth. Bryan Cantrill said it best: "You should open source your project because you never know when a magical little pony is gonna come flying by and shit out a fucking rainbow right in front of you.".

A siren server is the biggest and best computer on a network. Whoever has the most powerful computer would be the most powerful person, whether they plan to be or not

That makes no sense. It's not even wrong, it's a bunch of words strung together that don't mean anything.

In another article covering this story, I found this interesting quote:

"What we’re trying to do within a unit of my directorate [for Digital Innovation] is ... leverage what is becoming the instrumentation of the globe," [Deputy Director for Digital Innovation Andrew] Hallman added.


I welcome the CIA's attempt at modernizing. Let's face it: their recent track record at predicting major social movements has been terrible. The Arab spring, Tianamen square. They found out like the rest of us: cable TV and Internet. Monitoring social networks would seem like a good approach. I am wary of this technology being pointed domestically. Supposedly the U.S. is not in the CIAs purview, but there is probably no legal impediment to "lending" it to the FBI, or Homeland Security or whatever.

The article mentions deep learning as being one of the technologies that is being utilized here. I find it difficult to put all the pieces together on the efficacy of the system and the technologies being used as Deep learning is one of the most un-explanatory technology. That in turn raises some questions over the effectiveness of a system that predicts social and financial unrest. How did you do that?

Or you could just announce you have such a technology and thereby slightly inhibit the organization of uprisings via the Panopticon effect. All for the cost of sending a few emails.

Of course the reason what I say is unlikely is because actually building the system, functional or not, would justify an ever expanding budget.

I am not certain, but I think Recorded Future does something similar, and according to Wikipedia they are backed by the CIA. https://en.wikipedia.org/wiki/Recorded_Future

Something like this will become really useful when Google and Amazon are recording everything in US homes.

Cool. I wonder how they test this stuff.

Probably feed it with historical data and see if it can tell what has happened.

Backcasting is a fairly standard testing method, though you've got to be careful about overfitting.

I need to RTFA, but this sounds like it might have relations to Peter Turchin's work on cliodynamics.

RTFA won't do you much good. It's pretty light on detail. (IMO) It spends most of its time meandering about and trying to sound like some sort of really mind-blowing techno shaman.

So much so, that I doubt the authenticity and accuracy of the information that presented.

I felt like it read a little more like a paranoid fringe article.

I mean "Siren" in this case is named for the seductive nature of the power of the servers (greek myth reference) as opposed to a "warning siren" sounding the alarm of unrest.

It gets worse, rambling on about the danger (and inevitability) if/when this power falls into the hands of corporate america.


Yeah, I actually dug two layers further in, and tried to see if there was any info on the conf site itself (there isn't) on the talk / system. Overally, pretty content free.

Oh well.

Jaron Lanier coined the term. He's not entirely dismissable, but also not exactly tightly tuned to the heartbeat of the intel community AFAIU.

Dangers coming from usage of systems like this by a private corporation is not a cliche. It is a real danger.

If it has not clicked for you then advertising network is a behaviour control mechanism.

HN is a behaviour control mechanism. _Other people_ upvote articles and downvote anything they don't like. Therefore ensuring you are forced fed with only articles _they_ like. Your behavior is controlled by all these other people.


Sure it is and without sarcasm. Even when you do not post anything here then reading what is posted influences you toward certain target. Perhaps you get more interested in Javascript for example or most probably, startups.

The question is not in this. The question is in the scale of the operation and actual spread of the influence.

HN is big and people spend a decent amount of time on it.

Still it is minuscule compared to some real large scale operations.

What's the precision/recall?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact