In the field of audio forensics, Digital Media Authenticity (DMA) continues to gain importance. Current methods of authentication use the Electric Network Frequency (ENF). The ENF signal, present with varying fluctuations in all three US power grids, is embedded in just about any recording made. For authentication purposes, it is necessary to extract it from the recording in question and compare it to the raw ENF signal recorded directly from the power stations. Such a comparison can reveal a number of characteristics about the recording. One can determine, for example, whether the recording is an original or a copy, if it has been edited or recorded over, or if it contains starts or stops. What is the most efficient means of extracting the ENF? How does the extracted signal compare to the raw ENF? And what problems are encountered when conducting such research? For this paper, we propose to make several field recordings, extract the ENF from the recordings, and compare it to the raw signal. Depending upon the proximity of the recording medium to ENF power sources, the ENF will be embedded to varying degrees. As is such, we will collect a number of samples with varying ENF intensity. We will document both where and when these recordings were made to further explore the optimum conditions for capturing ENF in audio. Also, as closely as possible, our recordings will resemble those used in forensic analysis. A majority of audio forensic evidence contains voice recordings, and while ours will explore ENF extraction from a variety of audio samples, many of them will contain voice primarily with any additional environmental sounds. When analyzing the data, we will explore a number of ways of extracting the ENF, including minimizing DC offset, downsampling, and band pass filtering around 60hz. Throughout the examination, we will look at efficient means of ENF extraction, we will show the best conditions for making a recording with embedded ENF, and we will show what characteristics emerge when comparing embedded ENF with the raw power signal.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.