Slashdot Mirror
Audio Surveillance, Intended to Detect Gunshots, Can Pick Up Much More
New submitter groovethefish writes "This NYT article highlights the use of electronic listening devices installed on utility poles, buildings, and other structures, then centrally monitored for gunshots. The company SureSpotter claims it helps reduce time wasted by police searching for the source of gunfire in their patrol areas, but the privacy implications are just hitting the courts. If they are monitoring 24/7 and also pickup conversations along with gunshots, can that be used against the people who are recorded?" Evidently, Yes: the linked article describes just such a case. Continues groovethefish: "The company line, from the article: 'James G. Beldock, a vice president at ShotSpotter, said that the system was not intended to record anything except gunshots and that cases like New Bedford's were extremely rare. "There are people who perceive that these sensors are triggered by conversations, but that is just patently not true," he said. "They don't turn on unless they hear a gunshot."'"
QUOTE: "In at least one city, New Bedford, Mass., where sensors recorded a loud street argument that accompanied a fatal shooting last December, the system has raised questions about privacy and the reach of police surveillance, even in the service of reducing gun violence."
The Supreme Court has ruled people have no expectation of privacy in a public setting or publicly-open facility (like a mall). Note that also includes cops who try to make you turn-off your videocamera or audio recorder. They don't have any right to privacy either, and can not force you to turn them off, or confiscate & erase the evidence.
My AC stalker: " I personally agree with your posts most of the time, but that won't keep me from modding you troll"
"They don't turn on unless they hear a gunshot."
So, how are they listening for a gunshot, and then recording the gunshot, after the gunshot was fired?! Is that not a blatant lie or am I being daft?
Because they don't need to "be on" to "hear a gunshot"?
I think they mean the recording portion doesn't turn on unless the sensing portion detects a gunshot. A poorly worded sentence, to be sure. It's like your TV - even when your TV is "off", the small component that listens for your remote is still on.
How can you verify a sound was a gunshot THEN record it?
In my state of Wisconsin, it is against the law to record a conversation between two parties without the express knowledge of one of the parties. This instance would most likely be inadmissible in any court case. I believe this is the recording law in many states as well, but I only have experience dealing with it here.
If the only way you can accept an assertion is by faith, then you are conceding that it can't be taken on its own merits
I think a short circular buffer (a gunshot shouldn't take more than a second) is equivalent enough to not recording at all. If it's all in RAM, it's not permanent.
It still would need to record at least a buffer size big enough to go from the start of the gunshot to the time it takes to determine if it is a gunshot or not. It just doesn't save the recording unless it thinks it hears a gunshot.
I wonder if someone is going to go up to one and confess to their killings while shooting a gun continuously in the air.
Maybe, they don't record but use something like DSP analysis to trigger an event when an audio signature event is detected? Considering the audio is being "captured" and "analyzed" it does raise the question; is this equivalent to "recording?"
But why not have a buffer, I wonder? Say, 5 minutes worth? Then, when a gunshot is detected that 5 minutes of audio can be saved along with subsequent audio providing context around the shot. If no gunshot (or automobile backfire?) is detected, the start of the 5 minute memory buffer would simply disappear.
/me sips his coffee and ponders a new sig...
They don't have any right to privacy either, and can not force you to turn them off, or confiscate & erase the evidence.
They may not have any right to privacy but they certainly can, in real life, force you to turn them off, confiscate and erase the evidence. Doesn't mean it is legal for them to do so but they certainly are capable of doing it and probably will get away with it too. After all, once the evidence is deleted it becomes your word against theirs and they tend to hold the advantage there. Obviously cops should be held at least to the same standards as regular citizens (if not higher standard) but we know that it doesn't always work out the way it should in actual practice. The certainly aren't going to get thrown in jail and probably not even reprimanded and they know it.
I am tangentially aware of a military system that does the same thing and the way its engineered is you record to a time stamped ring buffer constantly. Meanwhile you analyze your ring buffer for a shot signature. IF you find a shot signature, then you perform a somewhat more detailed analysis to figure out the exact timestamp of firing (more or less). Then you uplink a really short data burst to central, something like "I'm sensor 23542542 and at 10:41:02.239582 I detected a shot and the local airtemp 73F and local air pressure is 1.0001 bar". Presumably central has a database of sensor locations, but if not a GPS RX on the sensor to generate timestamps works pretty well to report your presumably static location (although the .mil version I've heard about mounts on a movable APC).
Well anyway central optimistically gets about 10 reports, then its mega-triangulation time to pinpoint a location and estimated accuracy of fix.
Now if you dump the ring buffer to disk or something for possible later analysis, and the ring buffer is a minute or two (or an hour?) long, that's how you inadvertently collect street conversations.
This seems the only reasonable way to do this... any other way?
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
The TSA scanners didn't store images until we found out they stored images. Then we were told they only stored images for testing until we found out they stored images all the time. Then we found out the images were easily accessible to anyone after being reassured that there were ample security measures to prevent any yahoo from distributing humiliating or enticing images of some people.
I swear to God...I swear to God! That is NOT how you treat your human!
So, the device is actively recording and analysing all sounds, then when the last sound in it's buffer is marked as a gunshot it starts recording live.
"There are people who perceive that these sensors are triggered by conversations, but that is just patently not true".. so this is not quite true, these devices can hear your conversations, but unless you fire a gun, it won't store that conversation for very long.
It's the difference between "splitting hairs true" and "true for practical purposes"
Given a choice between outlawing guns and having a sensitive listening device on every street corner that can listen in on conversations like Big Brother, I'd prefer to outlaw guns.
Actually what he says is exactly true, the sensor is not triggered except by a gunshot (or presumably an equally loud and abrupt noise).
He didn't say that no audio is being recorded or listened for what-so-ever until a gunshot goes off (since that would be a self-invalidating statement)
While I agree it's not supposed to be permanent, a short circular buffer *is* a recording, and it means the device is on all the time.
For instance, the circular buffer used by the ReplayTV DVR for live TV pausing is supposed to be transient and inaccessible, but (due to a bug) it is possible to stream that video to other devices on the network. IIRC, the Tivo lets you save the pause buffer.
His statement shows that James G. Beldock is either ignorant of his company's own technology or attempting to "dumb down" the description of the technology to avoid scaring the common folk. Either way, it says nothing good about him or ShotSpotter.
TFA talks about an argument recorded by this system, associated with a shooting.
Unless the argument happened AFTER the shooting, it's unlikely they only start recording as a result of a shooting.
And it the argument DID happen after the shooting, I really can't see how it is even relevant to the case, unless the argument was on the order of "I can't believe you just shot that guy!! WTF?"
Which I don't believe, since I expect that if two people were wandering down the street, one pulled out a gun and shot someone else, then that two of them would be too busy running like hell to have an argument....
"I do not agree with what you say, but I will defend to the death your right to say it"
I implemented several IP cameras for a previous employer. They had a nifty little "record on motion" feature in which it wouldn't record (could still be monitored live) to the NAS unless there was motion detected. One of the big selling points around this feature was that it could actually record up to 30 seconds BEFORE the motion that triggered it so you could be sure you weren't missing anything before the camera was triggered. I believe it did it by keeping a couple minutes of video in a buffer on the camera. If an event triggered it, that buffer would be written to the NAS first and then it would continue appending the live video to the storage until the motion stopped plus X minutes. This system likely works in a similar fashion - it keeps a buffer that's continually overwritten until an event (IE: gunshot) triggers it to be saved to permanent media.
To continue my analogy above, the camera system could be set to only look for motion in certain parts of the frame. One of the cameras monitored the warehouse, so we could tell it to only trigger recording if there was motion at the warehouse door and to not trigger it if there was motion elsewhere in the frame. The gunshot would be the equivalent of the door and conversations would be the rest of the frame - it passes through the buffer but is never added to the storage medium unless it's occurring at the same time as the trigger. The camera system was pretty cheap too, around $300/camera as I recall.
How do they hear a gunshot if they're not already on? Maybe he was talking about the recording part of the sensor?
When I was there in 2005, some humvees had these. They didn't work well, and picked up a lot of false positives. When I returned a second time in 2006 / 2007, I don't recall seeing a single ShotSpotter. But the ShotSpotters made the guardshack's day when they could come back from dropping Marines at post and say "We were shot at! The ShotSpotter beeped, said it came from the right!"
I posted in more detail elsewhere in this article, but I installed a security camera system that could store some footage from before motion tripped the camera. Basically, once motion was in the frame (or a specific part of the frame), it writes everything from the buffer preceding the motion detection to storage and then appends the live video until X seconds/minutes after the motion stops. Unless there's a trigger, the preceding footage never gets written to storage. Technically, the buffer is a type of storage but it's very small, often overwritten, and only used if a trigger event is detected shortly thereafter.
You can do something where you continuously record and then throw away the recording after a few seconds if nothing was triggered. A lot of busses do that with video recordings so that they can have footage of a crash if it occurs.
DNA -- National Dyslexic Association
Two bamboo poles can create a crack that will set them off. A friend had the cops show up at his place in LA when the kids were playing with poles trying to injure themselves. The cops said that multiple gunshots were reported from there.
It is done on the DVR itself, not the cameras in our case. It continually records to memory, in an overwriting fashion, until motion is triggered. Then the data in memory is committed to disk, as well as what happens after that until the event ends (you set how long after it stops detecting motion that it keeps recording).
How about the police figure out why so many people are getting shot? Police show up after the fact and if drugs are involved the case goes to the bottom of the pile. Maybe someone could figure out why society has these issues in the first place? And don't tell me guns are the cause either because I seem to recall the UK having problems with gangs of knife wielding youths.
Only the State obtains its revenue by coercion. - Murray Rothbard
Well, it's the same as Google. Is GMail's mail scanning to determine ads to show you "reading" your mail? In a technical sense, yes it is, but in a practical sense, it isn't since it's being "read" or "recorded" by a purely mechanical device that cannot comprehend what is being said (for now).
As for the audio recording - it's a camera. The way the system works is it hears a gunshot, figures out where it came from (it's a microphone array), then pans/tilts/zooms the camera to that position to possibly catch the shooter on video. It may do some audio recording purely for identification purposes - to make sure it triggered on a gunshot and not say, a backfire in case the image is vague. For this you'd need a short backbuffer to provide context (and probably why the argument was recorded - they just happened to be arguing when a gunshot went off around them). Of course, if they were arguing and one pulled a gun and fired at the other...
The whole purpose is basically to cut down on calls of gunshots fired - because usually the caller is wrong, it's not gunshot but someone with a loud rapport. So being able to identify real gunshots from merely someone making a loud noise means not having to send someone out to investigate and maybe pre-call ambulances
So they can hear a gunshot when they're not turned on? And that's patently true? Jeez, have you americans invented magic and not told the rest of us? Yes, I'm sure Slashdot readers can think of loads of ways to do just that (pressure sensitive switches, etc., etc.) but I was more askance at the sheer mendacity of the spokesman. The case is only extremely rare if he is lying through his teeth.
The recent ruling that banned cops placing GPS trackers on suspects without a warrant was criticized by some of the justices themselves for not going far enough to clear up privacy issues in public. This is just another example of how those justices are being proven right, and at some point, the limits of what forms of warrantless, electronic surveillance (by private or public entities) can be used as evidence in court will need to be clarified once and for all. (I hope this happens sooner rather than later because the public will be more tolerant further down the road as they grow number to it over time.)
Ok, I had to read it twice but yes... the lawyer for one of the defendants in the New Bedford case is Mr. Camera.
Ascalante: Your bride is over 3,000 years old.
Kull: She told me she was 19!
Ok, obviously none of you guys are marksmen, or are forgetting ear protection tech. Let me give you an example which should clarify this and give further clarity to the "always recording" argument.
Guns are LOUD...a lot louder than they seem on tv...typically 140 - 190db. When there is a gunfight in a room and people have a conversation afterwards on tv, I chuckle. Unless they are wearing hearing protection (colloquially: "ears") all they would hear after the gunfight would be ringing. It is mandatory on most outdoor ranges and all indoor ranges that people wear hearing protection. Three or four .45ACP shots in an indoor range without protection, and you're deaf for at least the rest of the day.
Cheap hearing protection tends to be earplugs or big muffs, like you might see on an airport runway. But competitive marksman typically use electronic "ears" that permit normal conversation via an external mic on the headphones (and in fact amplify normal sound up to 45db), but shut down for any gunshot, essentially a fast audio gate. And they work flawlessly.
Once you turn them on, they are always listening for any sound above 85db, which they will attenuate by at least 30db, usually much more. How is that possible, if they have to "hear" the sound before they protect your ears from it? Because they are insanely fast, on the order of 1 - 1.5 ms.
I suspect similar technology is being used in the "gunshot detection systems" and it is very possible to only record the gunshot if the threshold is above 85db.
"The pie shall be cut in half and each man shall receive.....death. I'll eat the pie."
Alright, that would work.
So, why do I have a hard time believing that the police are using a system that just throws data away?
"I do not agree with what you say, but I will defend to the death your right to say it"
Can it discern the difference between say, an engine backfirin, transformer blowing or a person lighting off fireworks? .22 being fired off, there's probably a lot of other things that also sets it off.
I highly doubt that. If it has to be sensitive enough to pick up a
I am tangentially aware of a military system that does the same thing and the way its engineered is you record to a time stamped ring buffer constantly. Meanwhile you analyze your ring buffer for a shot signature. IF you find a shot signature, then you perform a somewhat more detailed analysis to figure out the exact timestamp of firing (more or less). Then you uplink a really short data burst to central, something like "I'm sensor 23542542 and at 10:41:02.239582 I detected a shot and the local airtemp 73F and local air pressure is 1.0001 bar". Presumably central has a database of sensor locations, but if not a GPS RX on the sensor to generate timestamps works pretty well to report your presumably static location (although the .mil version I've heard about mounts on a movable APC).
Well anyway central optimistically gets about 10 reports, then its mega-triangulation time to pinpoint a location and estimated accuracy of fix.
Now if you dump the ring buffer to disk or something for possible later analysis, and the ring buffer is a minute or two (or an hour?) long, that's how you inadvertently collect street conversations.
This seems the only reasonable way to do this... any other way?
Now if you dump the ring buffer to disk or something for possible later analysis, and the ring buffer is a minute or two (or an hour?) long, that's how you inadvertently collect street conversations.
This seems the only reasonable way to do this... any other way?
Sure, there's other ways that could be considered "reasonable". Since these are permanently mounted recorders, there's no reason why they have to record audio locally and discard old recordings. There's no reason why they can't have enough data bandwidth to let them stream the audio to central audio recorders that keep audio indefinitely (purportedly for further analysis or for "quality control" purposes). Likewise, storage is so cheap that even if data was stored locally and eventually overwritten, they could easily make the retention time days, weeks, or even years, so while it's technically true that recordings are eventually overwritten, they are able to be recovered for a long time.
Or he knows the equipment and when he says "recording" he's talking jargon about the device that is understood by the company the built the device and probably the people who work with the device on a daily basis. And it says nothing good or bad.
Already did it in a city called Canoas, in Brazil. Original source (portuguese): http://www.diariodecanoas.com.br/policia/301981/canoas-sistema-deteccao-de-tiros-passara-por-calibragem.html
The point is, that if they wanted to, altering the buffer circuit to continuously record is trivial. Its a software change, and so in that sense, they ARE recording at all times, because the option to exists.
Good-bye
To make a bad analogy, that's like saying that every time a cop draws a gun, it's the same as shooting someone just because the option exists and is trivial to implement. In reality, there's still a big difference between "can" and "does"
And in the case im talking about, all it takes is a proverbial one bit switch to go from limited record to always on, so one should assume they are ALWAYS on.
Good-bye
In Vancouver there are microphones on all stations and new trains on Skytrain (the above-ground subway system) that are being actively monitored by Translink Police. You can listen to the Translink radio communications on a number of free applications and hear them spot people sneaking alcohol on the train or talking about vandalizing ad posters for the cops. I think you may have to invent your own language these days, to even maintain an illusion of privacy anymore.
Bow before me, for I am root.
The military system is called Boomerang. It uses a cluster of microphones on a pole on a vehicle. The onboard systems can triangulate the shot directly, using the timing differences in the arrival of the gunshot.
When our name is on the back of your car, we're behind you all the way!
I can't find a link to it, but a few years back someone developed an IC that accurately detected gunshots, ignoring other noise. It was developed specifically for triangulation to determine the location of a gunshot. The IC in no way recorded audio info, but looked for gunshot signatures in the sounds it picked up. IF someone set up a system using that IC and let the IC trigger an audio recording circuit, then you don't even have the issue of a buffer.
I can mend the break of day, heal a broken heart, and provide temporary relief to nymphomaniacs.
So we're supposed to be worried about being heard by a microphone mounted on a utility pole in a densely populated area (I live in Milwaukee)? The white noise of the city drowns out any intelligibility of a normal conversation >10' away, and to accurately triangulate a gunshot's location the mics would need to be mounted somewhat high up (also to avoid being stolen).
Governments in the U.S. are EXTREMELY corrupt. Partly that's because it costs millions to run for office.
"A ______ is to a public microphone as a 1W laser from http://www.wickedlasers.com/ is to a traffic camera."
-- Terry
Didn't read the TFA again?
That's the entire point of the software.
Faster! Faster! Faster would be better!
These systems could be rather simple. Here is a guy who have made his own.
He states "Ohh gunshots, we love them actually in Lebanon :P gun shots on special occasions, gun shots when families fight, gun shots for civil strife. BIG gunshots when israel tries to f*ck with Hizbollah. Every one in Lebanon has a gun of some sort, or an RPG launcher. Oh well, the empty bullet casing should be our national symbol :P"
http://letsmakerobots.com/node/24544
I think they mean the recording portion doesn't turn on unless the sensing portion detects a gunshot. A poorly worded sentence, to be sure. It's like your TV - even when your TV is "off", the small component that listens for your remote is still on.
I think the 'saving' of the always-on recording doesn't get saved unless it 'hears' the sound it's supposed to. After all, how the heck is it supposed to record something after it happened? All this makes me wonder if more of these audio devices (and CCTV) will be destroyed/hunted by local criminals; if so, will that spur a new market for camouflaged devices?
No sig for you! Come back one year!
A well regulated militia being necessary to the security of a free state, the right of the people to keep and bear arms shall not be infringed.
I can't believe I had to invoke the 2nd, all the way down here at the bottom of the comments. Shameful, /. , just shameful.
But it possibly stands to reason that the deployment of such surveillance devices domestically is a clear violation of the 2nd Amendment, if not also the 4th Amendment .
The Admin and the Engineer
We should probably put a more critical eye towards possibilities.
The system turns on when it detects gunshots. But it's extra sensitive, so it often catches things that aren't actually gunshots. And it stays on for half an hour. Effectively, it's recording half the day.
Nominal behavior is not actual behavior.
Interestingly, this is also how your eyes work. If you briefly flash a bunch of words in front of people, and ask them to read back ones in certain areas, they can. If you want a few seconds before asking, they can't. Your eye dumps its visual data into a short-term memory buffer which fades after a few seconds. Any information not transferred from that buffer to long-term memory is lost.
Detecting impulsive noise, that correlates between sensors is enormously easier than detecting speech.
It's the difference between detecting a car parked on your foot, and a cat.
Linksys WVC54GA and WVC80N cameras will do this.
Michael J.
Root, God, what is difference?
Actually, this could probably be done without recording anything... just have a needle trigger hooked up to an audio compressor tube. If the needle jiggles too much, it trips the recording device.
Of course, in this case it would be impossible to tell if it was really a gunshot that set it off, as you'd never record the gunshot (or other loud noise) itself.
I thought Hong Kong had these back in 2057...
http://i.cubeupload.com/T6cyLu.png
A gunshot has a very wide frequency spread. Basically, take the Fourier transform of a delta function...
I imagine that there is a whole lot of overlap between the acoustic profiles of gunshots and other impulse-type sounds.
Ultimately, a gunshot and a backfire both sound like "impulse convolved with some transfer function depending on the local environment".
You could do it by listening for *two* impulses, or an impulse followed by some other noise (the thing getting hit)... dunno how hard that would be though.
If that's true, how do soldiers hear orders during a gunfight or right afterwards?
Hand guns are typically much louder than long guns. Additionally a solider on a battle field is typically outside where you won't get the echo that you do at an indoor gun range. Shooting a long gun in the woods or in a field while hunting is much quieter than shooting them at an outdoor range (roof overhead and concrete floor), which is still quieter than shooting at an indoor range. Also my uncle's handgun (.40 S&W) is much louder when we shoot cans outdoors than any of the rifles we have (Remington 770 .30-06, Marlin .30-30, Russian M91/30 7.62x54r). I suspect that the reason for this is the shorter barrels, larger diameter bullets, and additional unburned powder from handguns compared to rifles but don't know for sure.
Time to offend someone
systems were installed.
One of the very first systems was installed in a neighborhood I lived in for several years in California, which was simply inundated with "undocumented immigrants".
Prior to the installation, gunshots were such a common occurance in this neighborhood, that residents often laid-down on the floor of the homes in fear of stray bullets.
Healthy skepticism is a good thing, however the atypical slashdot knee-jerk reaction to this "news" obfuscates the reality that in places where these systems have been installed are markedly safer neighborhoods now precisely because of them.
It's depressing to see so many paranoid remarks from people that have no idea what they are talking about.