The speaker discusses the concept of using the mains hum, a sound produced by the alternating current (AC) running through power lines, as a form of watermarking or timestamping. This is because the frequency of the mains hum varies slightly over time due to the imperfections in the power transmission system.
The speaker mentions that engineers adjust the grid frequency to balance supply and demand, causing a tiny shift in the frequency. This shift is almost imperceptible but is present in the whole grid, regardless of location. The mains hum, which is a result of this frequency shift, can be heard in the background of audio recordings.
The speaker suggests that by logging the mains hum frequency every second, it could be used as a watermark to prove the time when a recording was made. This method, however, requires careful signal processing and a long recording for accurate results.
The speaker also mentions that the National Grid has released seven years of data, which could be used for this purpose. The data is processed to identify a component around 50 hertz or one of the harmonics, which is then compared against the logged mains hum frequencies.
The speaker shares some test cases, including footage from various sources. In some cases, the mains hum was clearly present and identifiable, while in others, it was not. The speaker emphasizes that the method is not foolproof and that it requires careful analysis and a long, uninterrupted audio recording for accurate results.
The speaker concludes by stating that while this method might sound like science fiction, it is a real-world application of technology.
1. The alternating current running through power lines changes direction constantly, completing a full cycle back and forth 50 times a second in Europe, but 60 times a second in some other parts of the world.
2. The transmission of this power isn't perfectly efficient, with imperfections and stray electromagnetic fields causing a small amount of power to become vibration, producing sound at frequencies of 50 hertz or multiples of 50.
3. This mains hum can be heard coming from power lines or speakers that aren't quite plugged in properly.
4. The grid frequency shifts slightly and constantly over time as engineers balance supply and demand.
5. The mains hum can be annoying for people recording audio, as it can be heard in the background sometimes.
6. If the grid frequency data is logged every second of every day and compared to recordings with mains hum, it could potentially be used as a watermark to prove the time when anything was recorded.
7. The National Grid has released data seven years recorded right at the source, which can be processed to see if there is a component around 50 hertz or one of the harmonics.
8. The longer the recording, the better for analysis. As a minimum, a recording of about a minute is suggested.
9. The main time frequency in the footage can be found and then code can be written to match against the millions of possibilities in the national grid logs.
10. The forensic team uses a signal processing algorithm to extract the peak and constrain that algorithm so that it can't suddenly jump up.
11. If a recording was taken on a certain day, then a day's worth of national grid data is given to the algorithm. If the exact time is unknown, all data is thrown at it, which takes a bit longer to process.
12. The team never said they were certain about the timing, only that they had a very high chance of being correct.
13. The algorithm was able to get to the second, although it's important to note that the team never said they were certain about the timing.
14. The algorithm was able to identify a strong component around 50 hertz, which was expected to behave in a certain way.
15. The algorithm was able to identify a strong component around 100 hertz, which was expected to behave in a certain way.
16. There was no strong component for the Electrical Network Frequency (ENF) in some footage.
17. In cases where there isn't a strong trace, the algorithm might not be confident enough to use for an analysis.
18. Sometimes, edits in the footage can cause a sudden jump in the NF frequency, which is quite likely to be an edit.
19. The algorithm can't work if there is a sudden jump in the NF frequency, which could be due to an edit or because the source of the ENF energy was walked away from.
20. Forensic analysis isn't a magic tool that always works. Compression can stop it from working, and uninterrupted audio of at least a minute is needed to have enough confidence in the result.