Contents
ENF analysis dates a recording from the hum of the electrical grid. Mains power runs at a nominal 50 or 60 hertz but drifts slightly from moment to moment, and that drift is identical everywhere on one synchronous grid at a given instant. A microphone near mains power captures the hum, and matching its tiny fluctuations against a continuous archive of the grid’s frequency can pin when the recording was made, sometimes to the second. It is the strongest and most court-accepted timing method in audio forensics, and also the most conditional, because it works only when the hum was captured. Cooper (2009) matched a 70-minute recording made in Glasgow to a London archive 676 kilometers away.
What is the Electric Network Frequency?
The Electric Network Frequency, or ENF, is the instantaneous frequency of the mains supply. Generators across a synchronous grid are locked together, so the small departures from the nominal 50 or 60 hertz happen in unison across the whole area, at every wall outlet at once. That shared wobble couples into a recording electromagnetically through nearby wiring, or acoustically through the hum of powered equipment, so even a battery device can pick it up. It is the only audio trace with its own statutory best-practice standard (ENFSI, 2009), and it has been used in casework since the mid-2000s (Grigoras, 2005).
How does dating work?
The captured ENF trace is a time series of frequency, and a national grid archive is another, recorded continuously. Dating is the search for where the questioned trace overlays the archive. The match can be astonishingly long-range, because one synchronous grid shares the same fluctuation everywhere: Cooper (2009) aligned his Glasgow recording to a London reference archive 676 kilometers away, and matched a two-minute extract against a 36-day archive. Where the alignment is clean, the read reaches court-accepted strength, which is why ENF is the most established timing method audio forensics has.
Where was it made: a grid, not a city
ENF also carries a weak location signal, but only to the level of a grid. From the fluctuation statistics alone, Hajj-Ahmad, Garg and Wu (2015) classified the grid of origin, continental Europe against the United Kingdom against the US-East interconnection, at about 81 to 88 percent. It cannot narrow further, because one synchronous area carries the same frequency at every outlet, so there is no within-grid geography to read. City-level work needs clean power-mains recordings at a signal-to-noise ratio above 20 decibels, while ordinary media audio sits nearer 9 to 15 decibels, too noisy to place, which keeps street-level ENF location at the research frontier.
What does ENF analysis need?
The method’s conditions are physical, not statistical, and missing any one of them breaks it. It needs the hum to have coupled into the recording at capture, roughly ten minutes or more of audio (two minutes at a good signal-to-noise ratio), and archive coverage of the right grid for the claimed period. Many recordings simply never captured a usable hum, and that is the ordinary case rather than a sign of tampering. When the trace is present, though, it is the richest timing evidence audio forensics has.
Can the hum be read or removed?
The hum is read on a spectrogram, as a faint horizontal line at 50 or 60 hertz with harmonics above it, extracted by a statutory protocol of DC removal, a narrow bandpass, decimation and spectral analysis (ENFSI, 2009). Removing it is not a clean escape: a notch deep enough to scrub the line leaves a spectral hole and inter-harmonic inconsistency that are themselves detectable (Chuang, Garg and Wu, 2013), so the absence reads as a removal rather than as innocent silence. Surfacing this and the other traces is what audio spectrogram analysis is for.
What it proves, and its limits
ENF answers when better than any other audio trace, and it contributes to whether, since a phase break in the hum flags an edit, a signal read alongside audio tampering detection. But it is still evidence, not proof: it is reported as a strength of support for a proposition (ENFSI, 2015), it needs reference archives and a preserved original, and it is silent whenever the hum was never captured. It sits inside the wider read of what forensics can learn from a recording.
Sources
- Cooper, A. J. (2009). The Electric Network Frequency (ENF) as an Aid to Authenticating Forensic Digital Audio Recordings: An Automated Approach. AES 33rd International Conference on Audio Forensics.
- Grigoras, C. (2005). Digital Audio Recording Analysis: The Electric Network Frequency Criterion. International Association for Forensic Phonetics and Acoustics.
- European Network of Forensic Science Institutes (2009). Best Practice Guidelines for ENF Analysis in Forensic Authentication of Digital Evidence (FSAAWG-BPM-ENF-001).
- Hajj-Ahmad, Garg, Wu (2015). ENF-Based Region-of-Recording Identification for Media Signals. IEEE Transactions on Information Forensics and Security 10(6):1125-1136. DOI: 10.1109/TIFS.2015.2398367
- Chuang, Garg, Wu (2013). Anti-Forensics and Countermeasures of Electrical Network Frequency Analysis. IEEE Transactions on Information Forensics and Security 8(12):2073-2088. DOI: 10.1109/TIFS.2013.2285515
- European Network of Forensic Science Institutes (2015). ENFSI Guideline for Evaluative Reporting in Forensic Science (STEOFRAE).