The Register has a fascinating story about a very new technique for forensic analysis of digital audio recordings. Called electrical network frequency analysis, or ENF, it relies on measurable variations in the the frequency of electric flowing through the grid. Digital recorders apparently now have such high time resolution and sensitivity that they pick up these variations. Even battery powered recorders may be affected, it being possible for grid frequency variations to be induced at a distance. Computer analysis of recordings can extract this distinctive fingerprint which can be used to help verify the integrity of a recording.
The Metropolitan Police’s digital forensics lab in Penge, south London, have been pioneering this technique which relies on a database of recorded variations over time. They have build up exactly that information for a wide region over a five year span, sampling once every 1.5 seconds. They also were able to verify over what distance the fingerprints match at any given time. For example, the article mentions they correlated ENF findings between London and Glasgow. Presumably mapping out how far this correlation covers, geographically, allows for more frugal monitoring and storage.
By matching recordings to their database, analysts can essentially time stamp the recording. They also can uncover covert editing, a cut-and-pasted recording will yield multiple matches in the database unlike a continuous, unaltered one. The British research and application builds on original work done by Romanian audio forensics researcher, Dr. Catalin Grigoras. The Register also confirms ENF has been used in at least one murder investigation though police wouldn’t identify the case for “operational reasons”.
This reminds me of my first readings of Van Eyck freaking in Cryptonomicon. It also makes me wonder what would be effecting in foiling ENF analysis. Power conditioning? Or would the conditioner introduce its own characteristic variations that could be matched to it, then from it to the grid?