The hidden background noise that can catch criminals
HTML-код
- Опубликовано: 5 фев 2025
- Electrical Network Frequency analysis, ENF analysis, matches background hum against power grid logs. I talked to one of the researchers who works on it, and also set them a challenge. Thanks to @answerinprogress, @hannahwitton and @SteveMould!
🟥 MORE FROM TOM: www.tomscott.com/
(you can find contact details and social links there too)
📰 WEEKLY NEWSLETTER with good stuff from the rest of the internet: www.tomscott.c...
❓ LATERAL, free weekly podcast: lateralcast.com/ / lateralcast
➕ TOM SCOTT PLUS: / tomscottplus
👥 THE TECHNICAL DIFFICULTIES: / techdif
It's really difficult to make a video about ENF analysis work visually, but I figured going out to a picturesque set of transmission towers might help!
3 weeks?
the
H o w
Tom, I have a question: Why do you always make your video public long after it’s been finished and uploaded?
We love you Tom!!! ❤️
Electric guitar players know this hum very well haha
Especially if you play a guitar with single coil pickups
Oh hi Rob
I'm starting to have this theory that every content creator watches Tom Scott. Rob Scallon - tick.
One time I accidentally used an AC power supply for a fuzz pedal and I became intimately familiar with the hum that day
Oh my God it's like all my RUclips dreams came true! Rob and Tom in the same thread? 😍
Or you can play to a slightly sharp G sub bass... Which never goes away... 🤔
It would be cool to see if people could disprove old hoax footage by finding cuts in the recording using this method. That would be awesome!
People have actually used this to find splices in faked speedruns, I guess that counts?
Maybe I'm thinking of only audio analysis, but I could swear I've heard that they used this specific type of audio analysis at some point.
Cuts can be found with just basic audio analysis, you don't need something this complex. Good luck finding the original footage, though.
@@nikkiofthevalley I remember a Super Mario 64 run gettind identified as spliced by using this method
@@nomercyformayhem2238 With ENF analysis? I know you can detect splices by looking at inconsistencies in the audio, but has specifically ENF analysis been used to timestamp mic audio and detect otherwise perfectly masked splices?
@@nomercyformayhem2238 That one was purely audio analysis, there were some obvious cuts in the audio that weren't noticable just listening.
(If I'm thinking of the right one)
If I ever need an alibi.. I just need to record a video and record some sound near lines in a different area... and put it in the video to clear myself. Thanks Tom for helping me with my future criminal endeavors
The grid frequency is the same nationally, you'd need to go abroad (and there's every chance that country would analyse the same... :) )
@@andylake7202 The time of day would be different thought wouldn't it?
Isn't that just make your record have 2 ENF?
@@swordsman1137 Yes but you can deliberately make the powerline hum much stronger, drowning the one you want to conceal
@@andylake7202
But if you can prove you were "in a different city" at the time of a crime -- landmarks show a different city, grid noise shows the time the crime happened -- it appears you couldn't have done it.
Tom is teaching us the best places to not go to get away with murder
Anywhere without mains power seemed like a good place even before knowing this though
Were you planning on recording the murder?
well just do it in the part of Yellowstone National Park that is in Idaho.
Secret Agent Style
All Class
And thereby the places to get away with it
Fun fact: old video game consoles produce a very similar sound to this and this method has become the easiest way to identify spliced/edited speedruns in the "completing games as fast as possible" community.
why not just say speedrun lmao
@@Daemon4 he did
I was wondering why so many of the radical left partake in speedrunning.
@@mozartl4866 what?
@@mozartl4866 because they’re good at detecting and calling out fraudulent people I assume?
As someone who works with audio, this is beyond fascinating.
You’d expect these kinds hums to just disappear and drown into background noise and become indiscernible.
The precision is incredible, to think you can obtain this information from a difference of less than 1 hertz in the quiet backgroud noise of a video.
Tom, I am a media forensics expert at the University of Colorado and I was just taken aback by you suddenly doing a video on something I am intimately familiar with. My instructor-now-colleague is the scientist that actually developed this method and published this all those years ago (in Romania at the time). Thank you for covering ENF like this. It is great to see topics I work with hitting popular videos like this. I would like to add, that there has been a lot of work relatively recently to get ENF data out of video, not just audio. Also, BTW, one of his earliest published papers on it was back in 2002... so almost 20 years ago now.
Way cool!!!!!
Wow, that's crazy.
I have so many questions. How can we run this code ourselves?
Need to open source this thing if it already isn't!
This is very cool, however, I assume that Someone could modify the ENF of a video as to make it appear that it was taken at a different time. Would that be detectable?
Its really interesting, and its one of those "Why didn't I think about that" things. Of course that variability would be a thing. gah
My master's thesis in 2013 was specifically on this subject: ENF signals and their use in forensic analysis. How nice to see this video being recommended out of the blue!
Must be nice!
Fantastic read, thanks for the info, Tarek!
thanks wat a great thesis
Do have the DOI for the paper for those of whom would be interested in reading? (:
Your master wrote an interesting thesis
Tom, I just have to say, I truly appreciate that you always put English captions. It really helps to understand what you're saying even though my volume is low. Thank you once again.
Agree ! Also helps when there's new vocabulary to me
Yes! Low volume, not the most familiar accent, etc. Captions are amazing!
Agree, totally.
Even makes it possible to watch in public without bothering other people
Agree
Note to self: Always add a layer or two with 50Hz from different days to any recording.
Or 60Hz, depending on the region
I would think that just slapping a high pass filter on your recording would work as well.
Or just say the date of recording in the video, so they can't get new informations
@@shawngeorge9194 what about the multiples of 50 Hz?
@@shawngeorge9194 gotta be careful of the 100hz harmonic, or the 150hz harmonic, or even the 200hz harmonic, how many harmonics before the noise floor becomes too high?
Cheers, tom! I’m a Criminal in the UK who uses US cameras spec’d at 60hz to confuse authorities who think I’m using 50hz. I also use low pass filters on my camera to eliminate all frequencies above 30hz and distort my audio to all hell
The fact you have informed them of your methods they can construct filters to scrub the data to unscramble your signals.
@@bighands69 nope, I’ve already thought of that and duplicated every audio track in audacity added random sine waves overlapping each other
Swiper no swiping! Did... did it work?
@@bighands69 You can't pull 1's back from a sea of flat zeros mate.
you can just remove the audio track at this point
If you look at this from another angle... we could create the perfect de-humming plug-in. You just provide the exact time and the plug-in finds the mains frequency and subtracts it from your recording -with zero "color".
more easily, you could filter out all frequencies from say 49.5 to 50.5 Hz, and perhaps the harmonics as well. But people will know you'd have something to hide.
@@tunateun well if you're sending a ransom video it's handy...
@@tunateun You can already do that Audacity. I've tried it on music, but that often has frequencies at the same level, and I think I can detect that when I listen to the filtered music. Would I notice if I didn't already know it was there? Maybe not. Maybe a pro musician or audio engineer would.
The mains frequency would likely not null with the one on the recording. There will be differences due to compression, the microphone's frequency response etc. but it would be interesting to see what would happen.
@@tunateun "people will know you'd have something to hide" unless you do it to every piece of recording
What I find funny is if this were done on CSI we'd be totally dismissing it as absurd Hollywood hocus pocus.
And yet I thought it was so clever when they isolated the noise of the el ("elevated train") in "The Fugitive" to figure out which city Richard Kimble was calling from.
Horatio: Ok, can you zoom in right there on their eye? Ok mag 3x. Can you zoom in there? Ah there is our killer.
I doubt we could do it here in the US with our lack of a national grid.
@@synthoelectro Enhance, enhance!
@@gamerballs oh right haha.
The best part of Tom Scott is he shares an example of it not working and dedicates time to explaining its limitations. Instead of just "here's a cool technique, what it can do, and how it works". Tom Scott does it right.
As a criminal myself, I am taking notes. Many thanks, Tom!
WHAT
BEANS
!!!
If this study made public, then your notes will help you yesterday.
NEED MOAR EDGE
From a privacy angle I have to admit I am not sure I 'love it' because I don't know how far it will be able to go in the future, but from the perspective of wanting to stop deepfakes/edited audio in the future from misrepresenting people, I am very much for it. The dilemmas these new pieces of tech create are extremely perplexing and impossible to pin down. Such an interesting piece!
And you're saying this while being online probably with little notion of how much of your information you're letting out. Also, if you're using whatsapp (like many) or any other social media (but especially whatsapp), the ENF hum is the least of your worries.
@@flybeep1661 True but I think there isn't an easy answer other than the present centralised net isn't good, what we do next seems to be unclear!
I would expect deep sound fakes to accidentally get this right
@@flybeep1661 You criticize society while being part of society. Curious!
Indeed. Anything a government can use to catch criminals it can also use to oppress its citizens, and the same goes for corporations.
never thought criminals might start using high-pass filters to not be caught
High pass wouldn't be enough since there are upper harmonics (100 Hz, 200 Hz etc) like in the second video example. A better one would be a mains hum filter that filters the 50 Hz peak and all the upper harmonics. It doesn't affect the sound that dramatically.
Or try and not be recorded at all as you won't have access to the original video to sound edit.
Talking of frequencies, Japan is an interesting country. Due to an unplanned event way back when, West Japan uses a power grid with 60Hz and East Japan 50Hz. This plays havoc with cinematographers/videographers like myself based in East Japan who have to remember to shift the shutter speed to 120 (or derivatives thereof) when filming indoors in cities like Hiroshima, Osaka, Kyoto etc to avoid flicker from the room lighting. Generally we remember. What we tend to forget is to put it back to multiples of 50 (usually a shutter speed of 100) when returning back to East Japan. Being careful to buy appliances that could operate in both areas used to be a thing, although these days most electronic devices, home appliances tend to be able to deal with both. Fridges and air conditioning units remain the most likely candidates for being either one or the other.
Whilst there is obviously NESW directions and even geographically named 'North Eastern Region' etc, Japan is divided up into a Western half and Eastern half. This is down to historical reasons more than geography. Much like the MIdlands in the UK. It is really only the midlands of England as 'mid' suggests halfway and looking at the island of Great Britain, the midlands are way down the bottom half and nowhere near the mid point of the island. Kind of like that, but a bit different.
I assume a shutter speed of 300 probably isn't worth it data wise just to solve that problem.
@@RAFMnBgaming It would make your footage very, very dark causing you to bump up the ISO which more often than not, introduces a lot of noise into your footage. I have used a ridiculous amount of shutter speed to counter the flicker from a projector when filming within a flight simulator. We somehow managed. - That was in East Japan, if you were wondering. :-)
I heard about this by one of the guys on trash taste, interesting
Same reason why when the eastern power grid was disrupted due to the big 2011 earthquake, the western grid couldn't supply the needed power without huge transformers.
@@seav80 Exactly!
Background sound analisys works with many different backgrounds.
During a thunderstorm, the claps of thunder are a precise timestamp, because lightning strikes are recorded worldwide for insurance purposes.
Near a railway station, the train noises provide a precise timestamp to any individual recording.
Thank you for the great video Mr. Scott...
Aircraft noise is another great example perhaps an even bigger one as planes cover more places and appear on recordings over a wide area. The sources themselves are fairly easily located (well over or near land anyway) as ATC secondary radar stores a timestamped log of every reply it gets from an aircrafts transponder. Oceans not so much as there is usually no aerial radar cover over the ocean, at least not unless there is a ship with an airborne surveillance radar running in range. That is if the data is shared even then but since most vessels so equipped are military they are often less than forthcoming.
@@seraphina985
ENF only requires an assumption of the grid being suspected and gives out a precise time.
Using thunderstorm requires a rough area and rough time. Best use to confirm area and precise time.
Using aircrafts requires precise area and rough time. Perfect for confirming precise area, time and date.
@@seraphina985 Aircraft time-stamping only works in proximity of a civilian airport...
I remember reading about a case where people recorded themselves committing a sex crime inside their home. They had a radio playing in the background of the video. Investigators used the audio to find the station and the exact date/time of the crime. Then, they narrowed it down further by listening closely to the static, and cross referencing that against the weather, and regional variations in the radio signal. Investigators narrowed it down to an area a few blocks wide. The perpetrators were caught.
INDEED
Fascinating. I have Just ordered a dozen fans, basins with little boats, and real-time compression algorithm coding for audio sources.
Now I can begin my heady empire of criminal pursuits in absolute secrecy.
😳📸
I’m actually thinking about the dual of this problem: grid operators can intentionally add signals to the grid frequency to create auditory timestamps. All grids need very precise frequency control, therefore, it might be possible to encode a signal there.
I see no immediate application for this, but this whole concept is amazing to me, as I’ve been working with energy planning for 15 years (so not power systems, but still related fields), and I’ve never heard about this before
There is plenty
For example, some kidnappers may kill the guy kidnapped, but release a video of him alive
In my country we had once trade captured dead soldiers for living terrorists, and this technology maybe could have changed that
Kind of similar to how economy 7 timed heating systems worked. The grid would send a signal on the mains and the switches would detect this and turn on storage heaters/water heaters. Some locations had similar systems for street light control.
Make it impossible to charge your phone without Google, Amazon and Facebook knowing exactly where and when you charged it. Assuming there is still detectable current alterations after the charges converter with the most suitable charger for the job and the highest possible current alterations.
@@freedomofspeech2867 The 50hz hum is a feature of AC (Alternating Current). A charger converts AC to DC (Direct Current). DC is not alternating so there is no 50hz frequency. There may be something left over (not sure), but the better the charger, the smoother the signal will be.
@@robhulluk The smartphone has a "phone" subsystem that includes a microphone. You just need to hope that "something" lets out a 50/100Hz noise. But... that is not really "when and where". The when is much clearer with the build in Real time clock of the phone. The where is just what grid, not a geographic location. When the grid spans wide like the whole of europe, then the location is as rough as "the european grid".
It can also be super annoying when you can hear it through the walls in buildings just because you can't not hear it. Means sleeping next to a plug is super difficult. I work in IT and regularly have to wear earplugs if I'm building up a room of particularly cheap equipment 😅
Yes, it Hertz just thinking about it.
...am I deaf what am I supposed to be hearing
@@thepinkestpigglet7529 often, it's not extremely audible unless you are near cheap or older equipment. Many homes in the UK/Europe are older than some parts of the world, so if you live in America or a Newer Apartment, a microphone would be able to pick it up, but it's likely you wouldn't notice it unless you are using some cheaper/sensitive tech
@@thepinkestpigglet7529 I don't hear it either
@@thepinkestpigglet7529 People usually lose the top end of their frequencies when losing hearing, not low end, you should be able to hear it if you plug in cheap equipment, I've personally heard it a LOT and VERY loudly when using some old 1$ mobile charger plugs plugged into mains power.
THIS HAS BEEN ON MY RECCOMENDED SINCE IT CAME OUT ITS LIKE IT HAS A CONSTANT SPOT ON MY RECCOMENDED
Dunno bout the poopy butthole person up there but it’s doing the same with my recommendations. I have the full red progress bar of it being fully played yet here I am again.
Oh wow, I had no idea this was possible and as a criminal, I am thankful that you raised awareness on this important matter!
Be careful now! Let's not get caught!
I can say with confidence, you typed this response 3 hours ago.
Same, I always listen to true crime podcasts to see how people get caught.
Next, Tom will be explaining how pollen forensics can trace the time and place of death of a corpse to within about 1 week, and to the exact town.
There's a British forensic scientist who has been doing this, putting war criminals from the Yugoslavian civil war in jail. Corpses alleged to have died in the mountains where they were being dug up were turning out to have come from miles away, at completely different times to what the "records" showed...
how does this work with pollen as the catchment area can be quite large. at least that is the approach people take with paleo/archaeological environmental data
There's a really good Royal Institution lecture on the - forensic anthropology, have a look if interested.
@@LordChesalot I believe its about ratios of different pollens found in the lungs... Different plants release pollens at different rates under given conditions, and its like a timestamp for the weather in a particular place.
Anyway, watching Tom's videos is going to give us all the means of a clean getaway. I'm wondering if a viewing record of the whole lot could start becoming incriminating in some way...
The Baden-Clay murder prosecution in Brisbane put forward evidence like this.
thats, definitely sketchy. There is no reason pollen would behave differently on a corpse vs a living person, so that CAN tell you something of value but it is FAR more likely to give you false results in most situations and its impossible to know which it will be giving you. That should not be trusted at all as a method of forensics beyond maybe vague guesses to look into.
I was a paramedic in NYC. Once I was doing an EKG tracing on a person on a train deep in the bowels of Penn Station. The EKG had a very weird, repetitive pattern on it, and I realized I was looking at 60 cycle interference, most likely from the third rail below us. It was very cool to see.
VERY happy to hear that Taha, Hannah, and Steve are off the streets and in jail! Thank you Tom!
The wobble of the mains hum is actually a fascinating topic on its own, as it is used to synchronize the whole grid and to match supply with demand. For exaple, if demand drops, the frequency will go up as the generators now feel less resistance, so they will spin faster. On the other hand, if there isn't enough power generated, the frequency will drop. This way, you know you have to let more water to the turbines or put more coal in the oven etc. Same as with a car motor whose RPM drop when the road gets steeper, causing you to give it more gas.
yes that is why in texas last year generators that were still going during the cold freeze had to disconnect because frequency was dropping too low. that puts massive load on generators and can destroy them.
I did not think I was going to understand what you were talking about but here I am glued
I never thought about this humming, interesting to see what's possible with it!
well, catching criminals
It seems like someone could use an imitation mains hum to encode low-bandwidth digital data. Falling frequency for zero, rising for one. At 60 Hz, a one minute kitty video could contain 90 words of hidden data.
@@hamletksquid2702 that's actually genius
indeed. totally true
This is very closely related to how some speedrun cheaters were found by some Japanese runners almost a decade ago. They found discontinuities in the mains hum of the "proof" videos that could only be explained by spliced footage. It ended up outing a bunch of people - including world records - in a few different games, but mostly mario 64.
Wild to see this pop up on RUclips as a forensic technique, but it kinda figures tbh speedrunning brings out the bigtime nerds.
Why is it that everytime I see someone talking about speedrunners it's about how someone was caught cheating? All these cheaters are giving that community a bad rep.
@@HOTD108_ Well do consider how it's often about other speedrunmers going above and beyond to call them out, and keep it fair.
@@HOTD108_ Someone beating a record doesn't really make the headlines because it's something that happens regularly
@@HOTD108_ To be fair think about all cheating in all other competition too like physical sport.
@@HOTD108_ because it's the thing that gets all the clicks that isn't "omg new mario WR".
Got it. Time to install those solarpanels with backup batteries and disconnect from the grid.
I am always amazed by your ability to bring us content that I never would have searched for and yet I always thoroughly enjoy!
Fun fact: Since cutting a video causes anomalies in the mains hum, countless cheaters have been caught with it in speedrunning.
Do you have any examples?
It has nothing to do with mains hums, audio of anything will have a distortion and abrupt transition in it if there's a splice. When detecting splices in audio there's no need for it to be mains, or even a hum. It can be music, or even the breathing of someone in the room
@@noahhounshel104 mains hum has also been used before. i remember summoning salt talks about it in one of his videos about mario 64 but i forget which
@@commanderleo they don’t compare it to a variation database! they just check if there are obvious splits or suspicious gaps
@@commanderleo They do? Any spectrograph can detect the mains hum, and you can find plenty of those for free. And you don't need access to these variations either to see the cuts, any badly made cuts will leave these distortions in the audio.
Now i finally know what it is that I hear constantly whenever i'm near any electrical device that is turned on. People around me never seem to hear it but I hear it almost 24/7. It's incredibly annoying and distracting at times, especially the times when it changes a lot.
Like a bzzz😂
It shouldn't really be audible, if you can hear an electric hum then something is not working correctly, either the device or the outlet it's plugged into
Did you hear the one they added into this video 25 sec in? I could always hear old tube TVs on as well.
Probably a neuro problem, unlikely you actually ear it 24/7 why lie about it
@@jenjen462those tvs were so annoying😂
This blew my mind honestly.
I already knew everything you spoke about but never imagined it could be used for this application.
Fascinating! Tom always delivers something you could go your whole life without knowing but are still delighted to find out about!
UwU
The mechanically generated audible hum is generally 100 hertz, or double the fundamental as force is proportional to the square of current, and squaring a sine wave doubles its frequency.
This is how speedrunners figure out if other speedruns recorded from CRTs have been spliced or not.
I’m a digital forensics officer who just started out. This was amazing to see. It always awes me how science moves so fast these days. Interesting video.
I thought of your job just yesterday, while watching an episode of Death In Paradise (cosy whodunnit series set in the Caribbean). There was a laptop they wanted examined, and one of the police officers said "I've got a cousin who's really good with computers", and just walked out with the laptop, with the apparent approval of her superiors!
I do not believe that this is admissible in court. nor would it be relevant other then to "show" a time that might not be correct. never mind the trust in your profession isn't exactly based in science rather authority and how you can make something appear to the jury. Actually backing this information up with a study that shows it's accurate (or multiple) will go a long way to sow trust in your profession.
@@TheObsesedAnimeFreaks hey I’ve only just started it!
@@laratheplanespotter good luck with your new career! How exciting.
@@TheObsesedAnimeFreaks I don't see why this wouldn't be admissible in court. Obviously, you're not gonna convict someone based on this alone. But if you know that something happened on day X between 9:00 and 12:00, this could definitely be used to say that "this video was probably taken at 10:15-10:23, which matches witness testimony".
All these different frequencies being pumped into us 24/7 - it's no wonder people are sick all the time, dropping from cancer and heart attacks left and right!
As someone studying Forensic Linguistics at the minute, this was fascinating! Keep it up!
Mm yes the area where the line between semantics and pragmatics is incredibly important. Are you gonna pursue a career in linguistics / forensic linguistics?
@@JonahNelson7 probably not but it’s interesting to study at undergrad level!
what is this i dont get it at all can u dumb it up,for me
UR Wearside Jack & I claim my £5!
Can anyone do this to find the timestamp for when Tom recorded this video? Reverse Uno!
Agh, you beat me to it! He's practically CHALLENGING his audience to do it!
Yes, done.
During the daytime, obviously.
I think the compression might give you issues
@@nahometesfay1112 Not just compression, but also other audio processing carried out in the editing process
@@firstname1lastname127 you're underestimating CGI tools
As someone who is a full time criminal, I find this very educational
Are you a lawyer?😂
I guess it partly depends on the electrical shielding of the audio path in each case, as you're potentially getting electromagnetic as well as acoustic components of the waveform
The only difference between the frequency history of EM mains interference and of acoustic hum is a tiny phase shift. Picking up a mixture of both components would just result in an equally easy to isolate waveform at a third phase
@@WillHirschUK My point was more that the degree of electrical shielding would affect the overall strength of the recovered signal, especially if the EM component was dominant
TIL:
- Record everything with a filter to remove anything below 150Hz
- Record everything with a clear 50Hz tone from a signal generator in the background
Both of these are great ways to make your audio sound terrible, but of course there are slightly more nuanced variations on these that should thwart ENF analysis for those determined to do so
Record just the mains hum. Record signal plus mains hum. Subtract mains hum from signal + mains hum to get just signal.
@@katrinabryce very funny idea. the problem is that you recording needs to in perfect sync, which it can not be. As part of the mains hum in the recording is electromagnetic and part of it is acoustic from "something else plugged into mains" you have various phase shifts. Also note that higher harmonics will have different phaseshift and amplitude modulation then the base frequency.
It is much easier to record as little mains hum as possible (humanly audible is good enough) and then add a random signature over it (barley audible).
You could just simply use a 60Hz generator in a 50Hz country or vica versa.Except if you live in Japan or a few other countries where both frequencies are standard.
Hey Tom. A blast from the past. Always enjoy your presentations! Stop in once in a while! Thanks!
Genuinely one of the most interesting things Tom has ever made, and I can’t put my finger on why.
Can you put it on when?
I'm not sure that it would sound like science fiction 20 years ago any more than it would today. Heck, at one point in the United States, we had clock systems in our schools that used an electromechanical means of regulating the individual clocks in our classrooms, where the clocks kept time based on mains frequency and there was a controller for dedicated electrical circuits that allowed the clock mechanisms to be slowed-down or sped-up to try to keep them synchronized with real time. When I was a child in the 1980s this system was still in use in one of the schools I attended, an older campus that must've had this system installed in the sixties before the school district switched over to using individual clocks regulated by quartz.
The technique is very interesting and I could see it being valuable in analyzing things like terror threats or ransom demands, or even confirming the veracity of evidence.
I think microwaves often use that. Mine does. If it would have a microsecond display you could see each day how well the energy was managed :D
I think Tom made a video about this very subject
What Tom doesn't explain in thus video but does in another is that network operators are very strict in maintaining the 50 Hz average over time, such that a day will always be 4320000 cycles such that clocks etc, while they might not be on the microsecond right, at least do not drift away over time.
@@IIVQ Theoretically yes, practicly it did not happen a few years back, so my microwave was a minute or two slower per half year.
@@steemlenn8797 Probably because some clocks are synchonised to the mains, and others have an internal crystal that runs at a nominal 32,768Hz as a time standard. The crystals are rarely set to exactly 32,768Hz and so drift relative to the mains.
OMG!! 😂 I paused this video at the end to leave a comment saying "We miss you, Tom!" The thing is, once I paused the video, I heard a very low, constant beep. I heard it throughout the video as well but thought it was your representation of the "noise on the main"!!! 😂😂
Perfect Alibi:
1- set up a recorder at home at the time of the crime to record the hum
2- go commit crime (not going into detail of how you can get cought by doing that) but leaving any internet connected devices or smart devices at home is always a good idea
3- record a video doing something somewhere else that's off the grid or just power your home off the grid temporarily
4- play hum recording while recording your video, not to mention Synchronizing time manually across devices to match alibi and of course no view of the outside or any kind of natural light while of course being off-line completely
5- present video as Alibi saying you were home at the time
6- ???
8- P̶r̶o̶f̶i̶t̶... high chance of still getting caught, depending on type crime and whatnot
The house I'm staying at has a disconnect to the power grid that allows me to switch to a generator when the power goes out, I imagine if someone else had a setup like that they could flip over to their generator and get away with whatever.
@bestplayerhear Not if the alternative is to make an honest living in the US during a scamdemic.
@bestplayerhear Occam strikes again :-)
Can you get away with arson like this? Asking for a friend btw
@@To-mos I mean.. it's not like ENF analysis is the only investigative or forensic tool available.
FWIW: the tone most people *think* of as "mains hum" is actually 100 or 120 Hz because it tends to be a non-polarized effect which doubles the frequency. For example, the power dissipated in a resistor is the square of the voltage and squaring a sin wave gives a frequency doubling effect.
Always wondered why transformers noise is not 50Hz
Also because the 50 Hz wave going through a full bridge rectifier results in a 100 Hz output.
"sign wave"
@@SnoFitzroy I considered that, but choose not to. I suspect more people would be confused by that spelling than the shortened spelling.
@@benjaminshropshire2900 It's a sine wave.
We had a lab extremely similiar to this in my aerospace undergrad. We took a recording in a room with white noise and 3 unknown instruments simultaneously. We used the fast Fourier transform to remove the noise and classify the instruments based on their frequency. Crazy to see the things we are learning applied to the real world.
Don't you just stand in front of the screen and say, "Enhance!"?
Fast and Fourier: a film about mathematicians racing cars
The way this works feels like an exploit in a video game that would be used by speed runners or high level competitors in an older game like Melee. Seriously cool stuff
Isn't this actually how spliced speedruns are detected?
It wouldn't have sounded like sci-fi twenty years ago. I learned about it in university fifteen years go, and it was relatively old-hat then. The technique has existed for as long as digital recording devices have been around, and it even works (with caveats) with analog (which have a tendency to vary their own recording rate dependent on the grid frequency which hides the signal).
well, twenty years ago is a bad time frame. 30 years ago it would have been sci-fi, because of the data processing and reliable audio recording.
Recording the data from the grid is just a problem that can be solved. (even with much older tech)
Comparing data requires a computer with enough storage.
Reliable audio recording means: when the tape motor is driven with 50 Hz mains, you would actually have to record a known constant frequency and use the apparent fluctuation to deduce the motor speed during recording.
Now with quartz crystals, that is easy.
@@sarowie Guess we can put that down to the "Wait, the 1980s are 40 years ago not 20" problem, and Tom's age.
I'm only just learning about it now and it still feels like SciFi. 20 years ago in 2002 we were only just getting usable digital cameras at accessible prices, and if you were to ask someone off of the street, not someone who went to uni to learn forensics, they would think it was SciFi.
Years ago, when everyone was searching for osama bin Laden, I had an idea, and sent an email to an FBI tip line.
I suggested that we could station radio transmitters at various locations in the vicinity of where osama was rumored to be hiding. When he would make one of his regular propaganda tapes, the recording could pick up on the various "radio pings", and through simple triangulation, he could be located. I never heard back. Still think about that to this day.
2:02 Achievement unlocked: Recognize Steve Mould from the top of his head only
Massive respect to Gully for disclaiming forensics during her bit, it's frankly a wobbly science in a lot of cases, but this was fascinating to see how it can be used!
I wonder if they could do something like... determine the distance from powerlines, and then plot a chart of all points in a city that are exactly that distance from powerlines, and then find where recording areas like houses intersect those distances from powerlines to come up with a list of like 50 addresses that the audio could've been recorded at.
Fun fact: Japan up to today has 2 grids: One with 50 Hertz (bought in Berlin around 1900) and one with 60 Hertz from the Edison company, bought at nearly the same time.
Maybe because of this (and Japan being THE producer of electronics half a century ago) we today have microwaves etc. that often work in both grids.
And the Japanese also use 100V, 120V, 220V and 230V nominal voltages (so everything in Japan has a voltage selector switch (or does it automatically).
I would've never thought this was a thing. Props to whoever conceptualized it/created it.
I knew this was a possibility from news stories in the past, but couldn't guess at what they would use to do the forensic work. Thank you!
The cool thing about this to me is that I already knew all the technical details but never would have thought it could be used as a hidden timecode.
Unless you are recording right next to a piece of electrical equipment, the biggest influence of main's hum will depend on if the device you are recording with is plugged into the an outlet. A battery powered recording device will be much less effected by main's hum compared to something plugged into a power outlet. The audio codec and encoder will also have a significant impact. I'd be curious to see this experiment conducted again when account for devices powered by main's power and devices that are battery powered.
The way this works is by recording the hum that is coming off electrically-powered devices, not necessarily the power of the device itself. Almost all devices that are powered by mains give off a hum including light bulbs, refrigerators, computer power supplies and a myriad of other devices.
@@josephteichman8102 ok thanks, so a terrible microphone wouldn't have enough resolution?
I saw this, when I am recording on my phone with a mic plugged and a charger plugged, there is a noise
@@JG-xm8jy microphones themselves are analogue so resolution doesn't come in to it,until the signal reaches the digital domain e.g a computer. A low quality microphone would most likely also have low quality cable and this would make a good antenna to pick up mains hum. A professional studio mic and cables would pick up less stray electrical noise i.e mains hum,but would have better audio quality.
@@andreiiosifescu that high pitched buzz,squealing noise would be from the switching power supply in the charger. Try recording and move phone/mic closer to charger, it should get louder. Then try it with other power supplies / chargers. If you have an old heavy charger (with transformer inside) plug that in and hold phone/mic near that whilst recording - see if you can pick up the mains hum.
Great walkthroughs. Well narrated. Concise and to the point; overall awesome video. Nice work.
I also learned this when I was studying electrical circuits.
Wires and small electronic devices produce a hum as well.
Interesting topic
@@LegendLength good one =)
Thankfully as I'm getting older I don't hear the electricity in the house as much [I've always lived in rural, quiet places.]
Its most noticeable when walking near electric pylons in foggy/misty weather.
The band Japan used that hum running in the background of their 1978 song "Transmission", which was on their debut album Adolescent Sex. It used to sound great played through my old valve amp stereo.
I love how they suck in an example of the hum at 0:23 - 0:28 it was a very nice touch.
One of the most insane videos I’ve seen from Tom. So much of the modern world is hidden in plane sight.
You can see quite a bit from 30k feet up
Despite it's limitations, this is a perfectly genius forensic tool that is also keep your privacy as it's an analysis of "background noise" against public records.
As a criminal, this is so true!
Interestingly enough, mains hums are also one of the things used as something to verify speedruns.
Question: If you did not have a recording of the EMF, but instead had multiple videos taken in the same place, around the same time, could you sync them up? For example, see if one video was 30 seconds sooner than another? This could allow analysis to determine an accurate sequence of events during a crime.
I'm not 100% sure, but I think you might be able to look for correlations across the entire audio spectrum to do the kind of alignment you're talking about.
Identifying a hum or constant noise peak that shows up as a distinct shape on a spectrum graph is the easiest way to do this, but I reckon you might even be able to do this with audio that the human ear (and a spectrum graph) would generally consider incoherent noise - looking at subtle shifts and changes in background noise, or shifts in echo etc, and then singling out and isolating as many discernible samples as possible, then correlating for them all in parallel. Would probably be trickier to present as evidence (an algorithm that could do this simply would be a feat, one that could be explained simply (in a law court) an order of magnitude more so) but potentially possible.
Recording engineers do it all the time in the studio.
If what Tom said is true, you'd need all of them to overlap for multiple minutes with at least one other, which would need so to, etc etc.
And they'd all need to have been close to electrical appliances for those minutes.
So, in theory yes, in practice no.
You would not be able to assume the delay between clips, but you could probably throw it into the code to search for a section in time that has all of the waves within a set boundary.
@@MrNicoJac You do not need as much footage if you know approximately when it happened, for example if you knew it happened in a certain hour, but not which minutes of that hour, then you could eliminate a most possibilities
tom scott is the master of entertaining and random af videos.
This is the audio equivalent of those scenes in CSI where they have a photo with a 3 pixel resolution and they say "Enhance" and they can resolve the image well enough for facial recognition.
Only this has the added bonus of being, ya know, real.
Can't wait for a CSI episode where they get a whole spectrum good enough to get the date from an object vibrating in the background of a photo.
The X-Files had an episode where Muldur believed an ancient potters wheel recorded ambient noises in pottery and accidentaly recording someone speaking.
Then there's the Fringe episode where Peter Bishop anylized plate glass from a window and extracted audio of a murder that imprinted on the glass.
As if Tom isn't capable of enough, he now knows how to fight crime with electromagnetic frequencies... Be afraid. Be very afraid.
Especially as government gets to decide what's a crime. Speech is a crime now. Self defense is a crime.
Lavender Jack?
Well, technically, ocular inspection was already a form of fighting crime using electromagnetic frequencies.
This is why every serious kidnapper has solar panels and a honda generator at the hideout. Basic op sec
Interesting question, how much of an effect does microphone and recording quality have on this analysis? I'd guess that a bad ADC or sample rate would kill your ability to resolve small changes in frequency. I'd also wonder if a nonlinear amplifier in the microphone could throw off the algorithm by mixing the frequency up or down a few Hz or distort the signal. This is probably well into Electroboom territory, but I'd love to see a colab between these two!
Electroboom is not so deep in to signal theory as you might think.
Dave Jones/Eevblog would be a better call.
A nonlinear amplifier does not affect frequency, just phase shift and amplitude, both with little relevancy.
Distortion also does not affect the frequency of the signal.
Mixing the frequencies does in fact happen - but that would only be relevant when two signals that are unrelated to mains happens to have a mix product in the very narrow frequency spectrum of interests. (which is not unlikely - videos are recorded with 50 Hz/60Hz to avoid light flickering, so multiples of crystal based 50/60Hz frequency are all-around the imaging sensor. But... those frequency are also the base of the audio recording, so those internal clocks signals are recorded as highly precise and stable)
I'd expect 4KHz 5 bit be bad enough to make this method ureliable on short recordings. But if you will master your records that way you will be to famous to hide.
@@sarowie You're a little bit deeper into signal theory than me but I'll try to keep up. We could be talking about different things when we say non linear amplifier. My understanding is that by definition non linear amplifiers affect the frequency in some way. I think I get what you're saying about mixing, but i was thinking that the second and third order mixing products would be the real problem. By my back of the envelope calculation, if you had even an 80 Hz signal signal somewhere in there you'd also see a 40 and 60 Hz intermodulation product. Maybe or maybe not enough to throw the algorithm off
Also check the microphone frequency response, if it drops off enough at low frequencies the ~50 Hz signal won't be picked up and you'll need to use another trace signal for this kind of analysis.
@@LadyAnuB Wrong, generally speaking, main's hum isn't picked up acoustically by the microphone itself, but rather by the electronic components and cables.
Good electronics will filter electric noise better than poor electronics.
Another factor, unbalanced signals are more prone to noise than balanced
Also, even if 50Hz isn't picked up, it's harmonics can be, 100Hz, 200Hz, etc.
With respect to "the team never said they were certain, only that they had a very high chance", this is a mark of a good forensic analyst and what you are legally required to do if giving expert witness analysis to a court. It's really easy when you start to do such analysis to make strong bold statements, however it's important to back off, and separate out what the facts show, from your analyst conclusions.
Dystopian Creep Factor: 88/100
Many audio recording devices have HPF (high pass filter) which you can enable to exclude low frequencies. This helps with removing mic rumble, hum and wind noises. The 50 or 60 Hz hum is usually removed by HPF, so this forensic method wouldn't work with HPF on. Though, one can also look at higher harmonics.
I'd like to take a moment to appreciate the fact that Tom seems to be friends with every other RUclipsr I watch.
Honestly I have more respect for someone who says, 'I have very chance of being correct,' than someone who says, 'I am correct.'
This is incredibly cool, good find Tom!
Wow! Tom, if you see this, thank you SO much for taking the time and effort to make these videos! I absolutely love them and I can confidently say that I have learned a lot from your videos! 😄 you rock!!
As you said it shifts and gestured a moving motion with your hand the birds in background were so perfectly timed
Tom Scott: Telling us all the things we never knew we actually wanted to know....
This is so fascinating to me.
As someone who finds harmonics, electricity, and frequencies to be quite interesting, I am very glad that Tom made this video. If you’re reading this, Tom, thank you.
And Tom, if you're not reading this, no thank you.
Finally the subtitles scroll at the same rate as I read
Tom teaches me more than my science teacher did, what a legend
A few years ago I've read an article about a couple that was on the run from the police. They uploaded a video and the investigators didn't only find out the exact time it was taken, but they also managed to narrow down the location thanks to a ripple control signal that was sent at same time as the video was taken.
WTF shoudl anybody upload a video if they are hunted by police? Didn't they know it? Or was it some sort of "give us money or... and here are the instructions"?
Tom, as a career criminal myself would you recommend I cut the power to the whole house before I go in and start streaming on RUclips from now on for my future victims to prevent capture?
In the good ol' days we had Omega causing interference instead. Used to live ~60km away from an omega station (Bratland), and the carrier waves (pattern of 10.2, 13.6 and 11.3 KHz) would show up in just about any electronic circuit, and if you where anywhere close to the antennae (like driving past it) you could hear it in your ear (no electronics required).
Reminds me of the GBR VLF signal that found its way onto Tubular Bells
What is Omega?
Sometimes my heating made noise. And by that I mean speach that I even was able to more or less understand 2 times even. Sounded like an announcer. But maybe a half-deaf neighbor had his TV on very loud and the sound transmitted through the pipes, not any radio signal. Who knows?
@@MottyGlix an old pre-gps radiolocation system
Instead of a constellation of satellites broadcasting a their unique timing signal on frequencies around 1GHz, you'd have a bunch of stations on land that broadcasted a cruder unique timing signal on frequencies around tens of kilohertz
By knowing which stations you could receive, and their relative timing, you could tell where in the world you were
The Russians still run their own version of Omega
Criminals be like: Thank you, Tom. We will not go to this spot.
The frequency variation is the same over the *whole* country. This might only apply to a tiny selection of criminals that record their work.
Only places far far away from mains and transmission lines and well shielded might not record something with detectable hum.
This forensic strategy couldn't be further from being "Beyond reasonable doubt". Then again, there is no justice.
Also, I appreciate the tips, next time recording myself committing a crime I'll make sure to turn my microwave on.
In the US, power generators adjust the frequency of AC over a day to ensure an average of 5184000 cycles per day. The frequency might not be exact minute by minute, but if there is drift it will be corrected later to ensure the accuracy of devices such as clocks that count the AC cycles.
I have to wonder whether, given a strong enough "hum" it might also be possible to determine a _very_ approximate location, i.e. North or south UK (or perhaps just uk vs Europe, where I think a previous video of Tom's taught us that although the grids are connected, they aren't synchronised, so the log could be different)
Oooh good point, I'm sure you're right that the logs are different from place to place
Synchronous grids tend to be very large, with only 3 covering the entire USA, and only 1 covering the entire UK. One separate one covers much of Europe. So the location wouldn't be at all specific, but you could tell what AC interconnection they were in, which wouldn't be particularly specific unless someone claimed to be in California but were in Texas instead. But for example the frequency deviations are the same in Boston and Florida. They'd also be the same in say Portugal and Greece along with most of Europe.
Makes you wonder if they will start adding audio hiccups to specific areas of the grid for the purpose of using for location if needed at a future time. I suspect the military has long since figured this out, especially with all the research into Havana Syndrome.
@@wildflower1397 isn't it the point that that's impossible? the entire grid is connected and electricity is fast enough that we call it instant
They hinted at that in the video when they talked about harmonics. Machines all have a harmonic signature and if they are near a large machine, think electric arc furnace, then one could in theory be able to pin point a persons location.
It’s similar to using a train whistle to tell how near or far away a train is.
This dude is like an entertaining teacher.
An interesting way to "fool" this would be to record onto cassette, then copy the cassette a couple times... Each time it's copied, there would be some introduction of mains hum, but at ever so slightly different frequencies
Then you also have the inaccuracies of the wow and flutter of the tape drives, which would probably make identifying it impossible as they'd be of a greater magnitude, though their frequency may be low enough you could compensate for it
Koenig, Bruce E. (1990-02-01). "Authentication of Forensic Audio Recordings". _Journal of the Audio Engineering Society_. Volume 38 issue 2. pp. 3-33.
It seems like some forms of background noise also foil it by masking the sound, as seen in the last video with the fan
You wouldn't even need that much effort - even being aware of the issue is quite enough. Several passes of triangular dithering will be an appropriate amount of anti-aliasing to blur distinct shifts, at least to the point of being defensible as natural occurrence
@@misterbigglesworth6320 nevermind that it would be really hard to prove the change in mains hum isn't from wow and flutter in the tape drive mechanism itself, and even if it's replayed on the same machine that recorded it it may not be in phase with how it got recorded. I think it would be a really hard case to prove!
If you actually wanted to fool this just use a 60Hz generator if you live in a 50Hz country or a 50Hz one if you live in 60Hz country.