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 …

This topic is closed for new posts.

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

While you are right in the main, you are wrong for this, I fear.

The electronics that do the frequency matching of the (say) PV panels to the mains will be fairly accurate, but it is very unlikely they will be so accurate that the micro-variations in frequency the plod are claiming they can measure accurately (and 100% ID a recording from) won't turn out to be substantially different.

I just can't see some Chinese engineer sitting there and saying "Yes, but we need it to be a perfect true sine wave running off this crystal oscillator" when honestly, a crude approximation of a sine wave will do - just look at nearly all the cheap inverters used in cars and vans.

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

"a crude approximation of a sine wave will do - just look at nearly all the cheap inverters used in cars and vans."

Yes but the ones used in cars and vans are not grid-tied inverters. The ones used in cars and vans can only be used **instead of**, not **as well as**, the One True Grid. The difference matters.

Suppose the grid is running at 50.000000 Hz and your proposed inverter is running at 50.01 Hz.

Suppose that at some point in time the grid input and the inverter output are in sync: same voltage output at the same instant in time. That's handy really, given that the two outputs are paralleled (that's what makes it a grid tied inverter, as used in solar PV or whatever, rather than a cheapo standalone job).

Suppose that some seconds pass. Now, your theory says the two can be at different frequencies. If (as you suggest) the frequencies are different, then the instantaneous values of the grid input and the inverter output are no longer the same voltage, even though they are both feeding the same piece of wire. In fact according to your theory there will come a point in time when one output will be (say) 120V+ and one will be 120V-. Do you see a problem there yet?

So, how do we stop them getting out of sync like that? We design the grid tied inverter so that its instantaneous output follows the instantaneous voltage fed in from the grid. In other words:

THE GRID AND THE INVERTER HAVE TO BE THE SAME FREQUENCY. EXACTLY.

Are folk getting the message yet?

but the noise is in the sky man!

it's. like, everywhere, like, all the time, man.

no matter how far off the grid you are!

solar, wind, biomass, batteries, or a bloke on an exercise bike.

you still get the hum.

Re: @2nd AC :-)

wiki SMPSU (Switched Mode Power Supply)

All that most kit sees that frequency dependent these days is 4 diodes (or a bridge), 2 big assed electrolytic capacitors, a fuse, a switch, and a beefy as you like chopper transistor.

with the loop impedance and the size of the (low esr) electrolytics, 50hz looks as close to 60hz as makes no difference.

most of the rest of it is DC, and the bits that aren't sort their own frequencies out (much much higher than transmission frequencies)

all part of UK national grid, therefore in phase.

Are folk getting the message yet?

isnt it funny when types that know 'digital' encounter the real world.

@ShelLuser

"Did you know that I have several electrical appliances which I can use both in Europe as well in the US ?"

No, I didn't know, but I'm hardly surprised. Any household that has electrical equipment bought in the last 20 years is likely to have such appliances. ...Which, I'm sure, the Met is well aware of...

All of that doesn't matter, though, as the whole point of such appliances is that they adapt to the grid at hand. They happily eat 50 Hz mains in Europe and equally happily eat 60 Hz in the US. So the buzz will be the local frequency. In fact, your appliances may well make life easier for the police, as at least the cheaper kind of these power supplies tend to spit out ferocious amounts of harmonics, making the buzz much easier to pick up.

I call bollocks on this

I have extreme doubts about this technique. As others have pointed out, the frequency of the AC mains is usually extremely close to the nominal amount, so you need very accurate frequency measurement to spot the differences.

Problem is, most home recordings are done through a soundcard (no shit, Sherlock?). Those soundcards typically get asked by the PC to sample at x samples per second and they deliver something like x - but their internal reference frequency is based on a nasty cheap ceramic resonator which is close to but not exactly the frequency asked. Maybe it's within 0.1% of the requested frequency, so instead of delivering 44100 samples a second it SAYS it's doing that but in a real second it's delivering 44050 instead. That goes into your file, you play that back through another sound card with a similarly off-reference oscillator and suddenly you have 0.2% (say) error in the actual frequency being replayed. A musician with perfect pitch might spot something very slightly wrong but to most people it will sound fine - just all the tones are ever so slightly off.

I first had this brought home in no uncertain terms trying to record off-air weather fax broadcasts, where accurate time sampling is crucial or your fax comes out all slanted. A cheap outboard SoundBlaster gives frankly hilarious results and you need to be able to tweak the alignment in software to fix it.

So, anything which takes 50 or 60Hz mains hum and thinks it can correlate that with something that was measured against a highly accurate refernce source but was actually recorded with something far worse - well, I look forward to seeing that stand up in court if the expert witness isn't a complete fuckwit.

Re: I call bollocks on this

I said that (but not so eloquently) in response to the 2010 article - regardless of what record/playback technology is used, where's the frequency reference that proves that 50.0Hz (or whatever) when recorded will be exactly 50.0Hz (or whatever) when played back? In the absence of such proof, all bets for this method's trustworthiness must surely be off.

Re: I call bollocks on this

Well, it may be difficult to prove when a recording was made. However, as the article states, it may still be useful to detect otherwise-undetectable edits. WHile the part of the audio you listen to might not reveal any obvious pops or clicks, the very faint hums of electrical interference in the background may well be jumping up and down like a yoyo.

I have to wonder if more determined fraudsters are using fourier-based noise removal to get around this?

Re: I call bollocks on this

I was sceptical about this when I heard it on R4 and I'm surprised to hear that one reference recording could cover the entire grid.

As far as accuracy of the recording frequency is concerned though, I assume they look for patterns in the relative change of the recorded hum as it drifts up and down over the course of the recording rather than matching to the reference hum's frequency exactly. So long as the recording's sample rate was consistent they could compensate for it being out a bit?

No not quite

Well the frequency error of your soundcard is in the range of less than 100 ppm. That's far better than 0.1%. Plus it's fairly consistent. It changes only very slowly. So yes, that introduces an error, but it's probably manageable.

Re: I call bollocks on this

i call your bollocks and raise you more bollocks!

you do not see pitch shift on recordings carried out this way.

you just don't

ever

maybe in the bad old days on tape, but even then only when the tape had been used a million times before and the recorder was on it's last legs.

i play and record on a vast array of kit of varying qualities and tuning just isn't an issue (i have pretty good pitch, but my SA is better)

Re: I call bollocks on this

the fft is how it works.

hence absolute time domain fidelity between recording and playback devices is irrelevant.

it started a X hz after 20 seconds went up to x+0.01 then after a further 30 seconds went to x-0.02

calculate the difference between the recording devices value for x, and the playbacks version of x, multiply out the time changes, et voila.

This post has been deleted by its author

Ah but

my gaming rig is Steam powered.

I think the point is ...

a massive PR campaign on behalf of the police to scare baddies into not challenging their "evidence".

Given police claims about cannabis factories being stuffed with UV heat lamps, I'm not sure we have much to learn from their science unit.

Isn't this only for recordings made by AC powered devices?

If I recorded something, I'd use my phone. I'd have to, as I don't own anything I could plug into the mic input on my PC. My phone, obviously, runs on battery.

I am highly skeptical that my recording would pick up a hum from the grid around me in my house and be audible to the extent you could tell a few hundredths of a Hertz variation for this type of forensic investigation. Or audible at all, to be honest.

Even if this worked when I made the recording in my house, what if I'm making the recording in my backyard or on the 14th green? If you're going to tell me I'll get an audible trace from nearby power lines, no matter how "nearby" they really are, I'll be forced to question your sanity.

I'd love to have more details on this, and know what made the recordings they were able to do this with, and where they were located at the time. I don't deny this is possible in the right circumstances, and is smart police work since keeping the record would cost nothing once it has been set up. But I doubt it is applicable very often - less often all the time as fewer grid powered recordings will be made these days.

Re: Isn't this only for recordings made by AC powered devices?

Well, possibly, though the way it's being sold is that it would work even if you were using a clockwork recorder on a train.

What matters with this sort of wheeze is not whether the technique is infallible or even moderately well-proven. What matters is how readily the courts accept it. The lesson of low-copy-number DNA 'matching', accepted instantly by the courts before anyone bothered to question whether it was reliable, made any statistical sense, or was properly understood by the judiciary might have been a useful one, but there's no sign that it's been learnt, and I strongly doubt this technique will be queried much, either. I suspect the first consequences will be that any recorded evidence supplied by defendants will be swifty declared iffy and that the Met will make a small killing for the privilege of doing so.

It's easier to disprove a hypothesis than to prove it*.

I think Burbage is on about the same thing that passed through my mind - all this could really be good for is proving that a given recording _wasn't_ made at a particular time because the variance in hum is too great, in the same way that DNA matching can only [conclusively] clear people because they don't match.

The risk of false positives is too high for a good defence lawyer to let the prosecution get away with saying that the 'hum profile' (or rather, the close approximation of it which can be extracted from recordings which almost certainly weren't made on kit designed to accurately preserve it) proves when a recording _was_ made. And of course if there's a gap in the pattern (aka doesn't match any known profile) then it's probably been fiddled with, and almost certainly won't be a 'kosher' recording made on kit connected to UK mains. But again, all that does is expand the list of known unknowns and potentially disprove a given claim, rather than being proof positive of any particular assertion.

*If you're a scientist or a logician. A lawyer is of course required to be neither on behalf of his client.

Re: Isn't this only for recordings made by AC powered devices?

it's not audible recording, the hum comes from the electromagnetic field produced by all those wires, the audible part comes from the nature of the recording machine - the field creates small eddy currents in the components of the recording device which are, to the device, indistinguishable from the signals coming from the microphone.

you clearly have a smart phone - get an app that reads field strength - search your app store for 'EMF', then go play golf and prepare yourself for a shock (figuratively, that is)

Re: Isn't this only for recordings made by AC powered devices?

if the recording medium was electronic, then a clockwork recorder would still have the hum.

the second para conflates science with statistics. saying one is 'sure' of something without sufficient evidence is always going to end in tears. - in this case, im sure the frequency record has already been sufficiently analysed to be able to say that 'we need a recording sample X seconds long to be able to have a significant probability of a match' as with DNA.

Re: Isn't this only for recordings made by AC powered devices?

you clearly have a smart phone - get an app that reads field strength - search your app store for 'EMF', then go play golf and prepare yourself for a shock (figuratively, that is)

Just because my phone's compass hardware can detect EMF fields doesn't mean that recording something with its microphone will result in a detectable 60 Hz hum on the recording. If you claim the microphone is affected by it, why isn't the camera?

I checked the iTunes store for "EMF" and the only app I saw that wasn't intended for detecting ghosts was rated only two stars :)

Re: Isn't this only for recordings made by AC powered devices?

the hum is NOT 'recorded' with the mic. it's NOT audio, its (as you point out) EM. its superimposed on the electronic signal derived from the audio signal from your mic.

the camera is - though sucking the data back out of a photo after jpeg compression would be CONSIDERABLY more complex.

AH! iTunes

nuff said, im an android man myself, plenty to choose from in our store, but then they let any old maniac write malware for droids apparently(who'd have thunk it)

Well that's a relief then, we can all sleep soundly now.

And in exactly how many hundreds or thousands of cases have criminals modified audio recordings to pervert the course of justice so far this century?

Re: Well that's a relief then, we can all sleep soundly now.

How many times have the police done so to edit a recording of an interview?

With this new technique the Met's forensic lab will be able to use the Met's archive to spot when the Met fiddles with a recording

Re: use the Met's archive to spot when the Met fiddles with a recording

Like that will ever happen (the cops spotting a 'fiddled with by the cops' recording).

Re: Well that's a relief then, we can all sleep soundly now.

I think the frequency of either the police or defendants doctoring the evidence is pretty low. But this technique is a tool with which the evidence can be validated.

The problem is; it appears that the baseline data is in the hands of the police. So they will be able to validate their recordings as required. But one can only hope that the data will be made available for the use of the accused as well.

Re: Well that's a relief then, we can all sleep soundly now.

"I think the frequency of either the police or defendants doctoring the evidence is pretty low"

Been ignoring the news this last few months?

Try telling that to the Hillsborough families, who have finally been able to see that two thirds of witness statements from the day were tampered with by the police, because they showed the police in a bad light, and who have finally been able to see that every subsequent inquiry till the present one has been a complete whitewash, aimed primarily at continuing the coverup of the incompetent police action on the day, and the disgraceful police tampering with evidence afterwards.

Re: Hillsborough

Written statements really aren't relevant to this discussion, are they?

Re: Written statements really aren't relevant to this discussion, are they?

In a way, they are.

If cops tamper with written statements, then it would be reasonable to expect some to tamper with electronic recordings.

It goes to the 'mentality' of the police (cover up our fuck ups at all cost).

I can easily believe this is sometimes able to prove a recording is fake if it was alleged to have been made at a very specific time (say it's a phone conversation and they have call logs), but I'm really doubtful the accuracy is good enough when the time window as more than a day or so. And of course proving a recording is real is much harder, since it could be tampered with.

But on the other hand I'm not really basing that on anything scientific.

If only...

"audio specialists at the Metropolitan Police forensic lab in south London have been recording the hum of this frequency continuously for the last seven years – 24 hours a day, 365 days a year..."

If only they hadn't taken two days off, the record would be complete...

I would think

You could use this to disprove someone saying a recording was made at a fixed date and time. If there is a miss match it's false.

But I don't think you can use it to say something was recorded at a given time for sure. How could you say that a given pattern didn't happen more then once?

Fitting the Crime.,..

to make the Punishment.

Scaring kiddies, that's all.

This *might* work if a kid uploaded a "mix" tape err, cd err, MP3. Maybe. But the kiddies aren't making money out of the technology. They are just sharing.

But I doubt it'd work , even there. Too much signal loss and cross-talk in the systems the kiddies are using.

Me, on the other hand, I have totally stable 60hz power (some old computer gear requires it). "Mains" is completely filtered out by the battery room ... If I was in the business of making money by selling other people's intellectual property, this newold technique would be utterly useless.

Methinks the Met Police's Dr. Alan Cooper is waffling to maintain a cushy job ...

This post has been deleted by its author

Re: Scaring kiddies, that's all.

Fuck me

MP3s? File sharing? Kiddies? intellectual property?

Do you actually read anything before deciding to post your ill thought out ramblings? Do you have any idea of what was being discussed in the article? Are you genuinely that hard of thinking?

And it looks like we have another item to add to the list of old shit jake owns... an entire room dedicated to batteries.

If you are going to accuse someone of waffling [1] then at least read what they are trying to tell you!

[1] for you to accuse anyone of waffling would be like the pope accusing someone of being catholic

Its even easier than Audacity

This is so dumb its not even funny.

"No your honour, I used this sound recorder optimized for voice dictation. I don't know why the police claims its false; all I did was click this "band filter" * switch here to get a better quality...".

* Every decent audio recorder which also provides support for better voice recordings knows a simple filtering technique such as "band pass filtering". Without going into too much detail this kind of filter picks up a specific range of the audio spectrum and, well, filters everything else out. My 'fail' comes from the mere fact that Its quite common to use band pass; so basically meaning filtering out low frequencies to get rid of unwanted noise, and filter out high frequencies to get rid of unwanted hisses and such. All at the same time.

In general the human voice sits between 80Hz and 1.2kHz (roughly). So a bandpass of, say, 150Hz to 2.5kHz wouldn't be an exception.

Happy rumble hunting!

Re: Its even easier than Audacity

how many db/octave is that filter?

Re: Its even easier than Audacity

What, pray tell, are you going to do about the induced noise at, say, 800Hz?

Another argument for alternative energy sources :)

Try matching that off-grid pattern..

fm

I share the skepticism, bit there is a difference between this and dendrochronology. Whereas there is no control over the climate vaiation that leads different tree rings, Plod could be getting the National Grid to introduce coded signals into the mains hum.

Re: fm

And backmasking our heavy metal records.

groundhog day? 1970's all over again

this technique was first mooted in the 1970's to verify recording location, authenticity etc.

well known was the fact that rubber belts on tape and record decks caused 'wobble' in music and voice

thoughts turned to sychronising tape and record decks with mains frequencies, but was found to vary so much, that quality of audio and early video recordings could only be guaranteed when referenced against crystal oscillators, used on some early direct drive systems.

so nothing new, just a reiteration of what we knew

(record - vinyl circular disks - for those whose parents were barely out of dipers at the time)

Re: groundhog day? 1970's all over again

I think if you look carefully at your keyboard you'll see that there are a couple of keys labelled "Shift".

Re: groundhog day? 1970's all over again

He's just being vewwy, vewwy quiet ..

Re: if you look carefully at your keyboard

Much better all lower case than too much upper case, no?

Re: groundhog day? 1970's all over again

Vewwy, vewwy, vewwy quiet?

Damn wabbit.

Who holds the reference data?

If you have recorded evidence of serious police misconduct, what is to stop the police from claiming its afaked because it doesn't match their reference?

What is to stop the police from altering recorded evidence and inserting the required hum from the reference data?

Page:

This topic is closed for new posts.