So if one were to set up a continuous recording of mains electricity, and then provide a UI that lets you grab a slice of the hum (a time range), and sell that, you could make some rather evil money. Or, if you were the police, have your "real" recordings verified very easily.
I don't think it's quite as easy as mixing in some of the correct hum in to your recordings. The real trick is to eliminate the authentic hum (which reveals the recording to be altered) and then substituting the correct hum. Really you would need to have someone with decent mixing/audio engineering experience to get this right.
If it's just the 50Hz frequency, it's easy to make a cheap application with a GUI that computes the FFT of a wav file, replaces the value at 50Hz using some straightforward parameters (e.g home/suburban street/city center) and IFFT back to an output wav file.
I may be wrong but this sounds quite easy to fake...
That would still sound tampered because you'd lose any bleed over from other (ambient) sounds. Having all bleed over suddenly stop at 50Hz and then a clean electrical hum would look very suspicious.
You can solve this pretty easily. First, record the victim, and tamper with the recording as needed. Call this tampered recording "sample A". You need this recording to be somewhere that ambient noise will not vary much throughout the day.
Next, go to the same room (or a similar room) at a time when the victim will not have a good alibi and when the room will have the same basic ambient noise. Record using the same recording device in the same conditions (e.g. if it was in your pocket for the original recording, keep it in your pocket for the second recording). This is "sample B".
Take your recordings into Audacity. On sample A, apply the "Equalization" effect like this: http://i.imgur.com/gxPTV.png
Now mix the two samples. As long as the ambient noise does not contain easily isolated components like other people's voices, you'll have a convincing forgery.
Now you might argue that there could be other forensics techniques to detect this kind of tampering, but I would argue that if such reliable alternative techniques were available, this mains analysis technique wouldn't be particularly valuable in the first place.
I don't think that's taking into account the harmonic frequencies. The 50Hz buzz is not a sine wave. That means that even with that filter, both 50 Hz mains sounds should still be detectable in the recording. Unless I'm missing something?
Ok, so you look up the true signal for the actual recording time, subtract that from the recording, and then add the signal from the time you want it to look like. Voila: perfect bleed over.
And its frequency is not constant. You'd have to have some kind of adaptive comb filter to remove the original buzz with all its components. I've a feeling that would be tricky to do in a way that doesn't scream "photoshopped" (or whatever the equivalent is for audio).
It's not even that complicated and you can do it with existing software. You can take the file into something like Audacity, band stop the range surrounding 50 Hz (or 60Hz in the US), and then simply mix in the prerecorded mains sound like you would do any other mixing.
It's not necessary to mix in the frequency domain because mixing in the frequency domain is mathematically equivalent to mixing in the time domain (FFT(x + y) = FFT(x) + FFT(y)).
This might leave ghosting of the original signal and would probably highlight any fluctuations inherent to "the main" that the phase cancellation would miss. I think this would probably be one of the cleaner and easier ways to proceed though.
Even more evil technique criminals could use: carry a device which emits its own fake, randomly fluctuating "mains" sound, thus tainting any legitimate recordings so that forensics analysis will incorrectly identify them as "fake".
Worse though. A bag of hair and skin cells could throw police off your trail. A tainted recording might could make it as far as the courtroom, only to have the defense bring in an expert witness to discredit it.
If you're not worried about the time; just proving the recording is continuous simply play it back through speakers and record from a microphone near the mains supply. The current mains noise will be significant enough to drown out the noise already in the recording, and you'll have your continuity for very little effort & no money.
Just thinking about it, couldn't you get an isolated recording of the hum, and then mix the inverse of it over the appropriate audio sections? That should leave you with a humless track.
- step 1: Set up a recorder in a relatively isolated environment, ensuring that the hum is being recorded.
- step 2: Record a 30 minute conversation with the target, ensuring that you have enough to splice together something incriminating.
- step 3: invert the hum recorded from step 1 and mix it into the track from step 2. This produces 30 minutes of humless audio with the target.
- step 4: edit the 30 minute recording to produce incriminating audio clip of about 30 seconds.
- step 5: overlay legitimate audio from 30 seconds of real hum onto faked recording.
It seems like it could work.
- step 6: Get caught for something much simpler that you overlooked, and ruin your life.
Please don't actually use this maliciously. I suppose it could be a decent defense in court if you could prove that it is possible to fake the hum.
If you've ever tried to do this you'll know that while it works in theory, getting it working in practise is much harder than you'd think. In the past I've tried to create acapellas using a song and its instrumental (both digital recordings taken from the same source) using this technique and even with the highest possible quality digital audio the results are mixed at best.
It wouldn't entirely eliminate the hum, but a balanced feed (note that XLR can be unbalanced as well) will reduce the EM interference and cross talk that adds to the background hum.
Another contributing factor is when power supplies to the equipment are different or "dirty". Often this can be resolved with something as simple as a multi-adapter with the earth pin disconnected (note, this shouldn't be attempted unless your hardware is already insulated. but then if you're trying to commit a crime, then an electric shock is likely preferable to life sentences in jail).
If the hum varies over time, I don't think you'll be able to perfectly invert it with a separate sample very easily. You'd probably have to simultaneously record it elsewhere (but close by) and you might still get artifacts--though those might end up missed in the overlaid track.
Probably doable, but a lot harder than your method.
Made me think, what about audio compression? Psychoacoustic algorithms might want to alter the signal, or cut off the some of the frequencies entirely. This could make the record neither valid nor invalid according to this method, since the hum would be too altered to be a valid information source.
I infer from the article that it's not a perfectly consistent background noise across the entire grid, but rather that it's consistent enough to get a match as long as the recording is long enough. As such, even if you did have access to their recordings, or made recording of your own, you still wouldn't be able to just invert it like that.
>which is, as the article states, a challenge even for recording professionals
Actually, it calls it an "annoyance", which is a good characterization of it. And that's for professionals trying to make very high quality recordings where they accurately capture the output of instruments within that frequency range while eliminating the mains interference in the same range.
If you're recording someone's voice on a portable recorder that's in your pocket or stashed somewhere in a room, you're not likely to capture much in terms of voice in that range in the first place. Even if you did, it would still be plausible that you didn't. So if you tampered with a recording and then replaced everything under 70 Hz with ambient noise recorded separately (including mains noise) in the same location, it would be pretty difficult to prove tampering.
A few minutes after posting that, I realized that a band-stop filter would work fine for spoken word. However, that might add a different set of tell-tales to the recording.
There must be a host of legitimate uses for the same data. The obvious ones:
1) Verifiy or correct the date/timestamps on recordings.
2) Geo-locate based on hum signature. Could apply to recordings or "live" conversations (e.g. skype).
3) Synchronising recordings.
"This buzz is an annoyance for sound engineers trying to make the highest quality recordings."
It hit me a while back when I was writing the software for a test stand for a hearing aid as a summer intern. Mysterious 60/120/180 Hz frequencies appeared on our analysis, soon to be discovered as the motors for the building's ventilation unit. It was barely noticeable to anyone, but it was quite obvious to our test equipment, even in an insulated box.
It may be that the reason was not the sound of the motors, but the effect of the motors on your electricity supply for the audio gear.
Isolating the electricity (running stuff from batteries) might give more benefit than the audio insulation of that box.
Not just audible, either. Electric motors are very naughty when it comes to interference, as are ballasts in light fixtures. I've seen a street light throw off enough interference to disconnect a poor guy's DSL every day when it turned on, and it was only after repeated troubleshooting and hassle that an eagle-eyed install tech noticed (by waiting with the customer at the right time). It's a legend within that company now.
OT, but I've seen this come up before and I wonder if HN could explain/justify this grammatical curiosity:
> "A gang were accused of selling weapons..."
"A gang" implies a singular entity (gang), but "were" is a pluralized use of was, as if "gang" was plural (as in, 'several gangs'). (I lack the vocabulary to properly articulate myself, since grammar is not my strong suit. I am probably not describing terms completely accurately). I've noticed this more and more in regard to singular forms of entities (typically compromised of many singular parts, such as corporations). For example: "Apple were..." or "Google have..." or "Microsoft are..." I notice that this seems to be more of a British English phenomenon.
My question is this: Why are people using what I will call pluralized modifiers on what I would consider singular nouns? What I would consider the "more correct" forms of the above examples are: "The designers at Apple were..." or "Google's board of directors have..." or "Employees of Microsoft are..."
Is this just a cultural clash between American and British grammatical conventions, or is there an elusive (to me) practical reason why one version is "better" than the other?
I apologize for thread jacking. Hopefully the more relevant comments will rise above this one.
A number of words like army, company, crowd, fleet, government, majority, mess,
number, pack, and party may refer either to a single entity or the members of the set
composing it. If the latter meaning is intended, the word (though singular in form)
may be treated as if it were a plural, in that it may take a plural verb and be
replaced with a plural pronoun: the government are considering their
position (alternatively the government is considering its position). See synesis.
As an American English speaker I find it delightful to read or hear occasionally. For me it's a nice, occasional reminder that groups are made up of individuals.
As a Brit who has lived in the US for a couple of years on and off it gets very confusing.
I can never remember which spelling is correct for the current situation for some terms: centre/center, tire/tyre, singular or plural collective nouns, etc. I know there are two variants, I just can't remember which one is the US one and which is the UK one; stuff like colour/color is easy, as is when to use sneakers/trainers (only made that mistake once) but some words play tricks on me.
Gang is plural, so in the King's English 'a gang were accused' and of birds 'a flock were flying' I've noticed that in various US sources a 'gang' is singular even though it is composed of many individuals, sort of like a corporation is a singular noun reflecting a collection of possibly many people. One of the things that makes English fun.
In "British" English, a collective entity (such as a gang, or a company) is generally referred to using pluralized modifiers. It's a cultural clash, there's no practical reason behind it.
Last time this was posted on HN, this question went unanswered: wouldn't variations in the AC line frequency be dominated by variations in recording speed? Even digital recordings have to be perfectly clock synced or they drift out of sync, and not necessarily monotonically. I've made digital recordings just minutes long that sounded fine to the ear alone, yet even after lining up the beginning and end, the middle was noticably mismatched.
Most modern electronic recording devices have little change in speed throughout the duration of the recording. Mechanical devices, not so much, what with wow, flutter, and unpredictable speed drift.
It is likely that what is being measured includes the shifts in the 50hz frequency from minute to minute, which would be less masked by a recording that was slightly off frequency.
It seems to me (and I don't have any formal qualifications to answer this questions) that computer clock drift is on the order of mega- or gigahertz, while this is on the order of 50hz - 5-10 orders of magnitude slower.
The speed of the clock that is drifting is unrelated to the rate or the amount of the drift. An audio sample rate of 48kHz may be driven by a 12MHz clock, and that 12MHz clock may exhibit thermally induced drift between, let's say, 11.99MHz and 12.01MHz. That will result in a sample rate drift between 47960Hz and 48040Hz. A perfect 50Hz tone recorded in those conditions will vary between 49.9583Hz and 50.0417Hz.
In the case I gave before, I was trying to synchronize a 48kHz USB audio interface recording with a 48kHz/30fps DV tape recording. If I lined up the beginning of the recordings, the ends were off by ~500ms (IIRC), which for a 2min clip means a 0.4% deviation. However, if I adjusted the speed of one of the clips to align the ends, the middle would be off by 500ms, suggesting a fluctuating deviation as high as 0.8% (if my middle-of-the-night mental estimation is correct).
According to [0], the UK grid is allowed to vary between 49.5Hz and 50.5Hz, or ±1%. Watching the meters at [1], [2], and [3], it looks like deviations of 0.2% are common. Depending on the frequency of the mains and recording rate deviations, it seems mains deviation could be swamped by the 0.4% variation I observed in real-world recording scenarios. Thus, I am skeptical of the forensic utility of mains frequency analysis, and would need to see evidence that forensic analysts are compensating for recording rate deviation, or arguments why it's irrelevant, before I would change my mind.
I think you are missing how much the spectrum of the drift matters. Just knowing the range over which the clock varies (say 5%) isn't enough. If the drift is slow - which thermally induced drift usually is, because it's driven by daily heating/cooling cycles - then it's possible to correct for it in the recording. Analysis techniques which are based on frequency-domain variations would tend to reject this type of slow drift automatically, but the details depend on the technique used. If, on the other hand, the noise on the clock is fast jitter rather than slow drift, things become much more difficult. On typical consumer recording devices, the noise floor due to jitter is way below the noise floor due to the limited SNR of the microphones and amplifiers used. The jitter noise is non-linear, which makes things harder, but doesn't tend to be a limiting factor.
It's been most of a decade since I worked on this stuff (in the context of radar and sonar), so that knowledge may be out of date. Still, I'd be surprised if clock (LO) jitter is the limiting factor in this type of analysis.
The baseline is 50 Hz, but how big are the actual variations? It's the magnitude of the variations that you need to compare against clock drift, not the baseline.
It sounds cool, but I wonder how reliable this method is. What's the false positive and false negative rate? Errors here would have real consequences -- for example, according to the article, it was used as a crucial piece of evidence in putting several men behind bars for decades.
So why is an article about the Metropolitan Police forensic lab in south London (UK) illustrated with a stock photo of what appears to be a US power socket? (It might be a Euro socket, but it's certainly not a UK one.) Do they not have electricity sockets anywhere at the BBC that they could have taken a picture of?
The mains frequency is an interesting topic. While it is true that it floats still its daily average is purposefully kept at quartz-like stability so that clocks can use it for synchronization. It also needs precise phase synchronization across the whole network (otherwise generators would blow up.)
I'm not so sure if the daily average is kept so stable actually (or maybe it depends on the country).
I worked at a power grid a long time ago, and I recall that in days where the grid had undersupply issues due to climate and high demand, the frequency was kept stable at 49.5-49.6 Hz for weeks or so.
The clock-sync is a nice wishlist feature; but the frequency affects power consumption which often may be more important or financially valuable.
In some of the power grids in the US, the clock-sync thing was something that the operations were required to maintain over a 24 hour period. One technique used to monitor this was a simple electric clock. Electric clocks have synchronous motors, meaning that changes in frequency of the supply would change the time. The target was to be sure that at, say, 6pm, that the clock would show exactly 6pm, even though it would vary a bit during the day.
Curious, is this "mains frequency" the same ringing I often hear when in alleged silence? And, no, I'm not talking about my hearing because I've had conversations with others who witnessed it :P - more like the ring old tube TV's used to make.
I believe that you should be able to hear overtones (2* or 4* mains freq) coming from lightbulbs and power supplies (computers, chargers, etc) in an average room - it could be tested by running a decent mic + solidstate recorder on batteries and checking how loud it is.
That may be true, but there are plenty of things in your home that serve that purpose. Right after reading the first sentence of the article, I started hearing it. Either 60hz or 120hz -- a low B natural. Maybe it's coming from the heating ducts, or maybe it's a fluorescent light bulb. Heck, maybe it's the fridge.
It's really bad whenever I plug speakers into the wall. Not sure if it's the mains, or the radio tower that's 500 feet away, which also "allows" me to listen to NPR at conversational volume without use of a radio.
The brain is designed to process signal and if there is none it will sometimes make up its own. I guess you can compare it to dark current in photo-multipliers.
People who have tinnitus will hear the sound you describe or something similar all the time. When I was little I had the same experience as you, but now I have developed tinnitus and never get rid of that sound any more. Sometimes it also becomes rhythmical which very annoying. The only thing that helps is to ignore it.
You're probably hearing sounds that your body is making. The sound of blood moving in your ears is pretty significant when you're in a silent environment.
The science-fiction series Babylon 5 used this as a plot device in one episode - a group of hostages were located partly by identifying the background hum of machinery in the terrorists' video of demands.
It can be defeated by jamming the relevant frequency bands, i.e. by overlaying other mains recordings with random fluctuations. No deed to remove the original.
I'm not sure a notch filter would be enough because, even though the mains is supposed to be a perfect sine wave, some harmonics occur (sometimes with a much higher frequency). Perhaps by filtering all harmonics as well? Sound quality isn't exactly a must in these situations.
It requires a technically astute criminal, though, and most aren't.
It requires a technically astute criminal, though, and most aren't.
Exactly. Many don't even use gloves. I was wondering on how precisely you could locate the hum and if you could deliberately create a hum signature for a region.
I'm a little disturbed that they are using this as evidence, yet one of the "forensic scientists" is quoted as saying
Normally this frequency, known as the mains frequency,
is about 50Hz," explains Dr Alan Cooper
That is incorrect. Normally it is exactly 50Hz. That's why the time displayed on battery-powered clocks drift over time, but clocks plugged into the wall stay correct.
It is not very exact. The total phase error can be hundreds of cycles.
In the synchronous grid of Continental Europe, the deviation between network phase time and UTC (based on International Atomic Time) is calculated at 08:00 each day in a control center in Switzerland. The target frequency is then adjusted by up to ±0.01 Hz (±0.02%) from 50 Hz as needed, to ensure a long-term frequency average of exactly 50 Hz × 60 sec × 60 min × 24 hours = 4,320,000 cycles per day.[21] In North America, whenever the error exceeds 10 seconds for the east, 3 seconds for Texas, or 2 seconds for the west, a correction of ±0.02 Hz (0.033%) is applied. Time error corrections start and end either on the hour or on the half hour.
The general premise is that there is an arbitrary, recoverable signal which has been convolved with the data signal. The generating function for this signal is a matter of record if you have access to the utility logs, but difficult to obtain otherwise without being physically at the time and place where the effect occurred.
I strongly suspect that it can be compromised under both of the following conditions:
1. You take recording A, then you take recording B at a time and place which you want to assert that recording A took place at. You recover the thumbprint from recording B, suppress the thumbprint in recording A, then apply the thumbprint from B to A.
This is not a trivial process, but you only really need a plausibly consistent result. A reasonably basic understanding of signal processing theory, a copy of MATLAB, and many pots of coffee should do the job. Then, you could automate most of it for the next guy.
2. You take many recordings at a series of locations of interest, while taking data about the power grid from nearby locations and from distribution nodes. You then attempt to predict the signal at a location from the characteristics of the surrounding area.
This is almost certainly possible, as generalization from distribution logs to the local effect is what makes their fingerprinting technique possible in the first place. It is not a trivial undertaking, and it's questionable how well it would be generalizable. But at the same time, it's largely a question of if you want the data badly enough to do the legwork, and whether you have a reasonably functional understanding of machine learning.
Regarding point 2, the article gave me the impression that they used historical measurements of the the mains frequency as part of the analysis. Since Britain is on a single grid, the local conditions shouldn't affect the recorded signal. I'm sure it is possible to find patterns (frequency drops slightly in the morning as the public utility tries to keep up with increasing demand), but you wouldn't know, for example that the utility overestimated demand on the particular morning that the audio was alleged to have been recorded (and thus that the frequency was actually higher).
As for point 1, there are a number of plausible angles to approach this, but I think that the forensic adversary has a huge advantage: Synchronous detection. The approximate time of the recording is known, as is the historical record of mains frequency. That allows for the possibility of huge processing gain, which might allow for recovery even after the amplitude of the mains hum is filtered to below the quanta of the audio system. I almost think you might have to Fourier Transform the whole audio record and zero out any component at f_mains +/- delta (and harmonics). That, of course, would look pretty suspicious to a forensic analyst.
Even that might not be enough if mains hum has a determinable effect on the data compression algorithm used to store the audio data.
Edit: The point I was trying to make in the first paragraph is that if you wanted to forge a recording, you'd need to have the grid frequency data. Having it for anywhere would be good enough and not having it for anywhere simply wouldn't.
Actually, it should be trivial to defeat without MATLAB:
1. Take the recording of interest
2. Use any audio processing software to apply a 400Hz-4kHz bandpass filter. If you don't have audio processing software or a computer, you can play back the recording over a phone.
3. Play back the recording via cheap computer speakers (no subwoofer) or over a telephone, and re-record
Low-end speakers don't have much output below 100Hz, so the new thumbprint from location (2) will replace the old thumbprint. The harmonics of the old will still be there, but bandpassing starting at 400Hz will cut out the strongest ones and make recovery of the original thumbprint very difficult.
> The general premise is that there is an arbitrary, recoverable signal which has been convolved with the data signal. The generating function for this signal is a matter of record if you have access to the utility logs, but difficult to obtain otherwise without being physically at the time and place where the effect occurred.
No and no. It's not convolved but mixed, and the signal is the same everywhere (on one grid), it's specific to time, not place, it's easy to obtain as long as you have the foresight to record it.
So yeah, somebody should record these fluctuations and put them up on the torrents for criminals and dirty cops to abuse :)
If the hum is inaudible to humans, but can be detected in the recordings, all that is necessary to create a plausible fake recording is to use a device that is incapable of capturing the hum, or severely compresses the hum so much that the hum is not distinguishable. This may not provide a means of proving the recording is authentic, but it prevents the defense from demonstrating that it was fabricated.
Hacking the system requires reproducing an identical "hum" and not necessarily compressing it. It seems difficult but I wonder if it is impossible for someone with a synthesizer and access to their database.
You don't need their database, you just need to record some background noise (assuming that you can replace the hum in your recording with another one).
interested in the false positive (hum is present) in which case the hum you need to "replace" something with needs to have signal correlations that match the ones in the database they've built up.
I imagine you could filter out the 48-52Hz range (probably less) and make your fake recording unverifiable.
Taking the audio from the time you want to claim the recording took place and adding it shouldn't be too difficult either, though that might leave other clues that it's not genuine.
Except you would get the harmonics at 120Hz, 180Hz, 240Hz, 300Hz, etc.; there are some harmonics that are going to be within voice frequency that you wouldn't be able to filter out easily without taking out some of the body of the voice. It would (at the least) be obvious that the signal had been tampered with if it were missing all of the 60Hz harmonics.
Another approach, dub the inverse wave of the real hum over your recording, which should (in my limited understanding) cancel the hum out, leaving a humless recording for a fraudulent hum to be overlayed.
> If millions of people suddenly switch on their kettle after watching their favourite soap, the demand for electricity may outstrip the supply, and the generators will pump out more electricity, and the frequency will go up.
Obviously, if a building is off the grid then there wouldn't be the same background hum, but what if the building has its own energy generation system (e.g., solar) and feeds back to the grid?
I wonder if the same technique could work for the microwave range and the cosmic microwave background. That is, is there a discernible "hum" in microwave signals from the CMB?
