back to article Police use 24/7 power grid recordings to spot doctored audio

Forging audio recordings is a lot harder than it used to be, thanks to a new method of authenticating recordings based on the buzz of the electrical power grid at the time they were recorded. The oscillations of alternating current (AC) produce a distinct frequency – 50Hz in the UK, 60Hz in North America – that varies slightly …

COMMENTS

This topic is closed for new posts.

Page:

Bronze badge
Boffin

Well there goes another one....

"In the kinds of amateur recordings that are often entered as evidence in court, however, typically no effort has been made to edit the power grid hum from the audio, which allows investigators to use the subtle variations in the frequency of the noise to determine whether the recording is genuine."

So the baddies start mixing in a cocktail of phased 50Hz (UK) synthesised in audacity or pure data at a level +10db on the real signal?

3
2
Silver badge

Re: Well there goes another one....

So the baddies start mixing in a cocktail of phased 50Hz (UK) synthesised in audacity or pure data at a level +10db on the real signal?

err, If the point is to use the mains hum to authenticate a recording, wouldn't finding a wobbly 50Hz superimposed on it immediately lead to suspicion of it not being authentic? If you're a serious baddy who wants to fake recordings, wouldn't it be better to make your recordings in a heavily shielded room, do your own continuous mains frequency monitoring and add your own hum from the time you want to fake?

5
1
Silver badge
Thumb Up

Re: Well there goes another one....

Indeed. I started to record the 50Hz signal when I first heard about it being used to verify cctv footage. Going to sell the recordings (incl. audio editing services) by the minute to anyone with sufficient funds. Sounds like a good business to me.

Anon for obvious reason

3
0
Silver badge

anon, I said!

Damnit

14
0
Silver badge

Re: Well there goes another one....

Don't forget for authenticity to supply the sounds of passing cars, trains, aircraft, etc.

Oh, that's 'BBC Sound Effects no. 57: Ransom demands and death threats'

1
0
Boffin

Re: Well there goes another one....

If one is doctoring a recording, do it while operating on batteries while on a boat in the middle of the ocean, thousands of miles from land.

1
0
Bronze badge

Re: Well there goes another one....

Except that the recording you're doctoring will already have the mains noise in it.

The trick, I think, would be to use a high pass filter tuned to 100hz - although even then you'd have to be careful for harmonics.

0
0
Silver badge

Re: Well there goes another one....

thereby exposing the tape as doctored.....

duh!

1
0
Boffin

Re: Well there goes another one....

You don't need to record it yourself, it's all recorder for you here.

0
0
Bronze badge

"recording the hum of this frequency"

Er, 'recording the frequency of this hum' seems a bit more likely.

5
0
Silver badge

Re: "recording the hum of this frequency"

Well either that or phase over time. Both is equally well.

0
0
Silver badge
Boffin

Heard this one on R4 yesterday...

and it did raise my skepticism level.

Without any data on the system, I'm guessing they're using something akin to the dendrochronology and matching a short segment - presumably by frequency - against the reference.

It strikes me that while this is possible in theory, it's going to need some pretty sophisticated DSP work to isolate the mains hum - in particular, since the mains frequency in the UK is so closely controlled, you're asking it to measure fractions of a tenth of a hertz from a device whose recording frequency control may be less than stellar. It's going to need lots of cycles...

Equally, the general pattern of grid power usage is likely to be similar at the same time of day on different days; you're going to be looking at very small timing differences to identify different days.

So while it's an interesting technique which *could* produce a timestamp, somehow I have the idea that it's being presented just as 'an expert says this is so' and thereby baffling the jury, rather than actually being as forensic as it is no doubt being presented.

If I'm wrong here, then I hereby tender my humble apologies to the expert witness... and I'd like very much to see the test data and the scientific papers which have no doubt been already published...

12
0
Anonymous Coward

Re: Heard this one on R4 yesterday...

While I'm not going to say the problems aren't real, this kind of technique is exceptionally easy to test. Take a thousand random recordings from the sample period, with a known but hidden date, run them through the system and see how many it gets right. Given how much DSP has come on in the last few years I'd be sorely disappointed if it was more than 1-2 errors.

2
0
Silver badge

Re: Heard this one on R4 yesterday...

I have the idea that it's being presented just as 'an expert says this is so' and thereby baffling the jury

I bet you're right. The theory sounds plausible, but to someone with a bit of actual experience and knowledge, it comes across as unlikely. It is clear that the recordings in question may often be amateurish and thus more hum prone, but by the same token they may be cheap digital with a brick wall cutoff below 100 or even higher. I note too that at least in North America, the frequency is extremely stable and the variations would not be within something which is likely significant in a badly recorded segment. One thing is for sure, now that this is talked about publicly, the concept can be more easily used by a professional forger, and work against successful law enforcement. A more sophisticated approach would be to record a hum sample in a distant place and at the time of the "incident" then use that in a later fake recording to "prove" when something took place.

This is a can of worms. Not to put them down, but my take is that this is just law enforcement trying to look more technically skilled than they are. Maybe this is even a bluff and this story is actually more related to social engineering.

6
0
Anonymous Coward

Re: Heard this one on R4 yesterday...

heard this one on the reg >2 years ago ...

http://www.theregister.co.uk/2010/06/01/enf_met_police/

9
0
Thumb Up

Re: heard this one on the reg >2 years ago

Me too! Thumbs up coward!

I also heard this on R4 and I thought to myself "That's that thingy I read about on the Register yonks ago!"

1
0

Re: Heard this one on R4 yesterday...

I think you are missing the difference between identifying the date and time from scratch, and verifying a date and time claimed by the witness. The latter is surely a lot easier as you are just comparing two signals for equality. Even if the witness time is approximate, you'll have a relatively small range of samples to compare against.

3
0
Bronze badge

Re: Heard this one on R4 yesterday...

Remember that most recording these days st to solid state devices. Wow and flutter are really a thing of the past.

I think as well as providing a form of timestamp, the system allows you to spot where a recording has been edited - sudden phase changes in the hum will point to something being cut out or inserted.

2
0
Silver badge

Re: Heard this one on R4 yesterday...

Remember that most recording these days st to solid state devices. Wow and flutter are really a thing of the past.

I think as well as providing a form of timestamp, the system allows you to spot where a recording has been edited - sudden phase changes in the hum will point to something being cut out or inserted.

I agree that wow and flutter are a thing of the past. That makes it all the easier to make accurate forgeries. It is also not beyond the means of most people to record with hum and other external electrical input below the noise floor if they have some expertise and wish to do that. There is no need to worry about phase changes in the hum if you add it at the end or avoid edits. Awareness of what the end result is supposed to look like is the key here.

0
0
Silver badge
FAIL

Re: Heard this one on R4 yesterday...

frequency is extremely stable

nope

as load varies (ppl using power) the the mech load on the generating sets changes, as more or less generating capacity is added/removed this also changes the loading, this causes stuff to spin up or down - F=ma and all that(we are talking very small amounts) - this is how demand is managed, control sees the frequency drop and brings more capacity online to meet it. the only other way is to.....

.....

well use a medium to tell you _exactly_what the load will be in 2 minutes time

anything that changes the relationship between the load and the generating capacity will change frequency - that's what makes a 'black start' so hard to accomplish*

*

black start - bringing everything back from a catastrophic failure of the grid - the large fluctuations in load as circuits are switched back in and generators brought online tends to cause other generation to drop outside of its frequency range and therefore automatically disconnect itself from the grid, this makes the original problem worse, knock out more generation and so on and so on till we are all in the dark again.

This is why it takes to long to recover from a cascade failure, and i don't think anyone with any knowledge of the history of T&D could claim that this is an unknown phenomenon in the US :-D in fact when i think cascade failure the first thing i think of is the US! they have had a couple of doozies over the years

As far as this technique is concerned AFAIK it's been around for donkeys years (well.. ish) and is well understood, and pretty much impossible to fake.

(clue - your master blackhat recording tech when adding in his 'fake' hum will end up with 3 hums!, the original, his fake, and the one laid down when he did the fakery. immediately exposing the recording as doctored, and therefore unreliable.

0
0
Silver badge

Re: Heard this one on R4 yesterday...

"As far as this technique is concerned AFAIK it's been around for donkeys years (well.. ish) and is well understood, and pretty much impossible to fake."

I'm wondering about some of the lower quality recorders out there. I had an MP3 player that did 8000Hz mono PCM with some freaky pseudo-compression (a-law?) that gave quite small file sizes. But the quality for anything outside of the range of speech was terrible. My phone has an app to record to MP3, usually 64kbit but can opt for 32kbit. Again, mostly "adequate" for speech but sucks at other things.

I'm just wondering if recordings made digitally with a system designed to attenuate/discard "unessential" information mightn't muck up a 50Hz signal too much to arrive at a reliable forensic result?

0
0
142
Go

Re: muck up a 50Hz signal too much to arrive at a reliable forensic result?

"I'm just wondering if recordings made digitally with a system designed to attenuate/discard "unessential" information mightn't muck up a 50Hz signal too much to arrive at a reliable forensic result?"

They'll be fine for any real world codec - you're not likely worried about 50Hz when doing this sort of analysis - that's a difficult frequency to work with, especially if you're going to be trying to verify things aurally. Of much more interest is the harmonics - 100, 150, etc, up into the low-kiloHertz (speech) range which (annoyingly for any sound engineer) tend to get picked up much more strongly than the fundamental. It's easier to pick out and analyze frequency changes in those ranges. In the event your compression algorithm for recording ditched the 200Hz range, then most likely 400Hz, or 450Hz will still be there, and you can derive the fundamental from there.

Remember how Shazam can almost without fail pick out a recording in the background of a noisy restaurant through a crappy phone mic - that achievement is orders of magnitude more difficult than what the Met has in mind.

You make a fair point though, as it is possible to envisage compression with the sort of frequency quantisation where this sort of analysis will not work (quantising every frequency to, say, the nearest 100Hz for example, as in a crude FFT), however that sort of codecs will make any speech sound completely robotic - think somewhere between Stephen Hawking and a Dalek.

2
0
Happy

Frank Zappa - Dynamo Hum!

Zappa finally has IT relevance!

http://www.justsomelyrics.com/763646/Frank-Zappa-Dynamo-Hum-Lyrics

http://www.youtube.com/watch?v=AOj41ex9ryg

4
0

Re: Frank Zappa - Dynamo Hum!

Legend.

That is all.

1
0
Silver badge

That's Dinah-Moe Humm, heathens! (was: Re: Frank Zappa - Dynamo Hum!)

"I poked'n stroked till my wrist got numb

An' you know I heard some Dinah-Moe Humm

Dinah-Moe Humm"

--Frank Zappa, December 21, 1940 – December 4, 1993. RIP

0
1
Silver badge

Familiar?

Isn't this the same technique reported two years ago by el Reg : Met lab claims 'biggest breakthrough since Watergate'?

5
0
Anonymous Coward

Re: Isn't this the same technique reported two years ago by el Reg?

It certainly looks that way (thank you, I remembered that article, but nothing by which to uniquely find it quickly).

And comments on that article reference papers published in 2003.

Still, it's always good to see the Met are on the ball.

0
0
Silver badge

But it DOES get complicated...

Here in the USA where there are basically three power grids (West, East, and Texas) the frequency doesn't change that much. That being said, the power companies make a special effort to have exactly 5,184,000 cycles of power each and every day (save leap seconds), so clocks won't accumulate errors. The power companies take considerable pride in their 60Hz they generate, and a sample every second or so is probably differentiated by the error in the sampling equipment, most likely due to the ambient noise.

The bigger trick would be to have a recording and try to determine when it was recorded, talk about a needle in a haystack. Who knows if it will be successful.

The other problem is when recordings are made "off grid", battery powered and away from the hum (out in the country?). Not going to get much power line grunge here. Compound that with an AC inverter run off a car battery which generates much better quality hash, and you won't be able to do much.

Now where are those yellow stripes that are on my laser/inkjet printer.........

4
0
Bronze badge
Holmes

@Herby - Re: But it DOES get complicated...

The frequency is very stable in the UK too. Forget about the wow that someone mentioned here. Having said that, the UK grid no longer attempts to get an exact number of cycles each day, as there are so few synchronous clocks anymore.

It is also the case that a mainstream recording studio will have kit that is extremely good at filtering mains hum.

I think the point is that pirates will not have such good kit, so their mains hum will be detectable, with frequency variation more typical of an East European or Eastern country perhaps. To compare with the original they have copied, it will not be looking for a "needle in a haystack" - the prosecuting recording company will presumably have a log of when recordings were made, and then the pirate copy could be shown to have a hum variation different from that on either the master recording (if it can be detected) or from that as monitored by police at that time.

Presumably also, if the recording companies get keen on this, they themselves could record the main frequency at the same time as the music recording, and keep it on record (to use a phrase).

0
0
Silver badge

Re: But it DOES get complicated...

out in the country?

really, if you are standing 'out in the country' and cant see a pylon or pole line, chances are you are standing on top of a cable!

I don't know for sure, but i would expect there are _very_ few places on earth entirely free of mains derived emf.

0
0
Anonymous Coward

Re: @Herby - But it DOES get complicated...

It's not about pirates, it's about evidence

"I was in the pub with me mates when he was done over, the cctv and karaoke recordings will prove it"

type of crime

remember. most guilty people have an alibi before the crime

0
0
Silver badge
Happy

So take a high-pass 200Hz filter and ...

place it in the record path.

See:

Rene Christensen, "Active All-Pass Crossover Networks with Equal Resistors and Equal Capacitors", Journal of the Audio Engineering Society, Vol. 54, No. 1/2, 2006 January/February, pp. 45-53.

Alternatively, after all editing has been completed, simply apply a 50 Hz tone and keep Plod amused.

Might be interesting to use on the Watergate tapes with the 18 minute gap.

1
0
FAIL

Re: So take a high-pass 200Hz filter and ...

Usual problem. If you've taken steps to remove the mains hum, it points towards a doctored recording.

1
0
Bronze badge

Re: So take a high-pass 200Hz filter and ...

@This Side Up

Hardly! Many bits of recording equipment have this built in because it is easy and cheap, and gets rid of annoying mains hum without issue. Anything with a mains filter automatically takes that step. Also, anything digital voice recorder-wise is probably only recording between 62Hz and 2kHz to save bandwidth and improve sound quality.

0
1
Silver badge

Re: So take a high-pass 200Hz filter and ...

Also micrphones used close to the mouth typically have a high-pass filter since when you are close to the microphone there will be more bass in your voice.

So it makes sense for anything below 200 Hz to be cut off.

Of course if you record the mains independently you can just remove it and replace it with different mains hum and it would be near impossible to find that out.

0
0
Silver badge

Re: So take a high-pass 200Hz filter and ...

depends on the filter!

a bog standard butterworth 3db/octave jobbie, or simml'a would see the mains hum down by 6 db, or 25% of what it was to start with.

zillionthth order chebyshev, prolly down by a lot more :-)

but the purpose of this type of analysis is not so much identifying hookie copies of commercial recordings, but establishing the veracity and continuity of covert recordings. the record can show that _this_ is when the recording was made, and _that_ is what the defendant said - un-edited. If the record is corrupted or contains any anomalies then it is just another piece of contaminated evidence and will be chucked out faster than OJ's sock.

0
0
Silver badge

Re: So take a high-pass 200Hz filter and ...

DOH!

the hum is everywhere, any ferrous metal will pick it up.

quoting 2 numbers for a filter is almost meaningless

we are looking at signals way below the s/n threshold anyways

all the filter does is _reduce_ the signal level below a threshold

reduce != remove

0
1
142
Boffin

Re: So take a high-pass 200Hz filter and ...

Errr... numbers aren't necessarily meaningly - a 48dB/Octave HPF 200Hz filter will push any noise at 50Hz well below the level encodable on a 16bit system.

But anyway, that's moot point - as I've just mentioned in a different post here - of much more interest when doing this sort of analysis are the harmonics in the 500Hz-3kHz range - not the fundamental. It's utterly impossible to remove the higher harmonics of mains hum without mangling the speech.

0
0
Bronze badge
Big Brother

Re: So take a high-pass 200Hz filter and ...

Seemed to me from the R4 proggy that this technique used on phone call recordings too. Now these already have at least a 200Hz filter on them (heard much bass on a phone call lately?).

I think the system relies on harmonics being detectable (an imperceptible 'buzz') as much as the fundamental freq.

Now, back to my evil lair to fire up the portable generator, before issuing the ransom demands...

0
0
Big Brother

What struck me is this. How can they tell what hum is what?

Let's have a look at my own home computer. It's connected to a cheap UPS, is connected to the television and the sound system. Now imagine all those slightly out of phase 60 Hz frequencies all fudging together and trying to figure out what is what? How about taking into consideration the slight differences in frequency that might arrise in the CPU as I record ? Or how about a cheap recorder that doesn't work exactly 100% correctly as far as recording actual frequencies. Not enough to be distinguishable by human ears, but enough that they might not even keep it accuratly compared to themselves. Think cheap clock that isn't 100% stable in a digital recorder.

That's more like trying to tell who died and when by looking at the blood patterns in sea waves if you ask me. Or maybe they just believe the bullshit the guy with the fat commission told them.

2
0
Bronze badge

What cheap clocks that aren't stable? It's not a significant engineering problem to create a clock that keeps time. Even if time did slip, the phase change in the mains hum from a doctored recording would be easily detectable.

I had a DAT walkman 20 years ago that produced perfect recordings using a tape mechanism. Recording to solid state storage accurately today is trivial.

0
0
Joke

"What cheap clocks that aren't stable?"

Maybe the Cheap POS clock that's in my microwave, loses at least a minute a week!

1
0
Bronze badge
Holmes

@ Mad Chaz

Wrote :- "What struck me is this. How can they tell what hum is what?...Let's have a look at my own home computer. It's connected to a cheap UPS, is connected to the television and the sound system. Now imagine all those slightly out of phase 60 Hz frequencies all fudging together and trying to figure out what is what?"

They don't need to figure out what is what. They only need to show that the mains hum on the authentic copy (if detectable) is different from that on the pirated copy (if present).

0
0
Silver badge

mains out of phase??

you getting the power for your telly from a different grid (that is, as good as saying continent!) to the one supplying your computer - neat trick that.

(hint, no you aren't, you just misunderstand the premise - besides all the kit you mention would have SMPSU's and that a whole other can of worms)

I think you'd need a rather expensive recorder to pitch shift on the fly as opposed to a cheap one.

you think that slight variations in clock speed in a clock running at 100's or more probably 1000's of Mhz will have a lot of influence over a 50/60Hz hum??? (not my understanding of the behavior of crystals - a few cycles per billion if theres a massive temperature shift, but pretty stable otherwise)

0
0
Anonymous Coward

There are already holes in this. More and more people are getting off the grid. So if they are producing their own power, the buzz at home would be different than elsewhere. You also have where people might have solar, wind, etc. as a supplement which would also change the buzz since they are basing it on load.

4
2
Anonymous Coward

"as a supplement "

Look mate, I'm far from convinced by this Metroploditan BS either, but the thing is, it's based around the frequency of the power. If you have local generation from wind, solar, etc **as a supplement to** (ie at the same time as) the incoming grid, that local generation's output by definition must be working at exactly the same frequency as the incoming grid (go read about grid tied inverters, for example). Otherwise the wiring would be on fire (or at best, fuses would be blowing and either the local generation or the incoming grid would go offline).

3
1
Silver badge

@2nd AC :-)

You're right about the frequency being usually the same as the one on the grid. However... Did you know that I have several electrical appliances which I can use both in Europe as well in the US ?

Honestly; its not as dramatic as you make it. Most certainly not a totally impossible scenario.

In fact; this could easily be another interesting excuse for a court: "But I use EU appliances from a friend...".

0
3
Anonymous Coward

Re: @2nd AC :-) re "same frequency"

I don't know what you're on about, and it's not clear you do with your 50Hz vs 60Hz reference.

Where a grid tied inverter is in use, the incoming grid, and the inverter output, have to match *exactly*, all the time. In voltage, frequency, and phase. Otherwise fuses blow, or worse.

Doesn't matter (in principle) whether it's a nominal 50Hz setup or a 60Hz setup.

3
1
Coat

Won't work for large areas of Scotland. We have hydrodynamic power grids fed from water and wind power....

0
1

No you don't. You have many hydrodynamic generators feeding The One True Grid through grid-tied inverters and other frequency locking mechanisms.

The only way you'd have different frequency variations is if those were on their own completely isolated distribution system. That might be the case on some of the islands but the mainland is all connected (as far as I know).

3
0

Page:

This topic is closed for new posts.

Forums