Sonic Watermarking Method for Ensuring the Integrity of Audio Recordings

摘要

Methods for inspecting the integrity of audio recordings become a necessity. The evolution of technology allowed the manufacturing of small, performant, recording devices and significantly decreased the difficulty of audio editing. Any person that participates in a conversation can secretly record it, obtaining their own version of the audio captured using their personal device. The recordings can be easily edited afterwards to change the meaning of the message. The challenge is to prove if recordings were tampered with or not. A reliable solution for this was the highly acclaimed Electrical Network Frequency (ENF) criterion. Newer recording devices are built to avoid picking up the electrical network signal because, from the audio content point of view, it represents noise. Thus, the classic ENF criterion becomes less effective for recordings made with newer devices. The paper describes a novel sonic watermarking (i.e., the watermark is acoustically summed with the dialogue) solution, based on an ambient sound that can be easily controlled and is not suspicious to listeners: the ticking of a clock. This signal is used as a masker for frequency-swept (chirp) signals that are used to encode the ENF and embed it into all the recordings made in a room. The ENF embedded using the proposed watermark solution can be extracted and checked at any later moment to determine if a recording has been tampered with, thus allowing the use of the ENF criterion principles in checking the recordings made with newer devices. The experiments highlight that the method offers very good results in ordinary real-world conditions.