Forensic audio : the relevance of Electric Network Frequency Analysis as an authentication tool in South Africa

Abstract

The predominance of digital media in modern society has presented a need for methods of authenticating digital audio. When an audio recording is presented as part of legal or business proceedings, it may be necessary to provide proof of authenticity. This minidissertation investigates to what extent Electric Network Frequency (ENF) analysis can be implemented as an authentication technique within a South African context. The electricity supply frequency exhibits random deviations in time from the nominal frequency of 50 Hz. ENF analysis is based on the premise that the supply frequency is detectable in audio recordings made on equipment utilising, or in close proximity to, electrical equipment that utilises the main electrical network. The recorded ENF signal can be extracted and used as a unique temporal identifier for determining the authenticity of an audio recording. Results of experiments conducted in this study indicate that it is possible for a detectable ENF signal to be induced into an audio recording, and subsequently isolated for further analysis. It is possible to compare the extracted ENF signal to the extracted ENF signal from a reliable database to determine temporal authenticity, and investigate the integrity of the file. It is found that the following minimum requirements must be met for successful ENF authenticity analysis: • The recording device be powered by the main electrical network, • The equipment be proven to be suitable for use in ENF analysis, i.e. no UPS device is used; the microphone does not have any electrical components that could alter the ENF signal, and the sound card does not induce any frequency components or noise into the signal; and • The availability of a reliable comparative temporal database of ENF values. Furthermore, results indicate that establishing an ENF temporal database by recording the supply frequency utilising a data acquisition unit, and subsequently processing the signal, is more practical than recording the database as an audio file. It was found that the South African electricity supply is sufficiently random to be suitable for ENF authenticity analysis. Recommendations for further study are made.