Electric Network Frequency Based Audio Forensics Using Convolutional Neural Networks

Abstract

Digital media forensics can exploit the electric network frequency of audio signals to detect tampering. However, current electric network based audio forensic schemes are limited by their inability to obtain concurrent electric network frequency reference datasets from power grids. In addition, most forensic algorithms do not provide high detection precision in adverse signal-to-noise conditions.This chapter proposes an automated electric network frequency based audio forensic scheme that monitors abrupt mutations of tampered frames and discontinuities in the variations of electric network frequency features. Specifically, the scheme utilizes the multiple signal classification, Hilbert linear prediction and Welch algorithms to extract electric network frequency features from audio signals; the extracted features are passed to a convolutional neural network classifier to detect audio tampering. The negative effects of low signal-to-noise ratios on electric network frequency extraction are addressed by employing extra low-rank filtering that removes voice activity and noise interference. Simulation results demonstrate that the proposed scheme provides better audio tampering detection accuracy compared with a benchmark method, especially under adverse signal-to-noise conditions.

1D-CNN-based Audio Tampering Detection Using ENF Signals

“…The experimental results demonstrate that there is no proportional enhancement in the model's performance with an increase 14/19 …”

“…The study conducted experiments on audio data tampering detection by inserting with 1-second, 2-second, and 3-second segments, whose results showed the highest detection accuracy for 3-second insert tampering. Mao et al 19 proposed a two-dimensional convolutional neural network model for binary classification of original audio and tampered audio. Zeng et al 27 proposed an audio tampering detection method based on ENF phase sequence representation learning.…”

“…The advent of deep convolutional neural networks provides a compelling alternative, enabling the automatic extraction of latent features from audio without enabling the. Mao et al put forward a two-dimensional convolutional neural network for binary classification of raw and tampered audio 19 . However, their approach focused solely on the fundamental ENF wave, and neglected the higher-order harmonic components, thereby omitting some distinctive characteristics.…”

ASLNet: An Encoder-Decoder Architecture for Audio Splicing Detection and Localization

“…Lin and Kang [8] proposed a wavelet-filtered ENF signal to highlight the abnormal ENF variations and employed autoregressive coefficients to train the classifier under a supervised-learning framework. Mao et al [9] utilized the multiple ENF features as input eigenvectors to the convolutional neural networks for detecting spliced audio. Meng et al [4] used the spectral entropy method to determine the length of each syllable and calculated the variance of the background noise of each syllable, then judged whether there is an operation of the heterogeneous splicing tampering in the audio by comparing the similarities between the variance of the background noise of each syllable.…”

“…However, when the signal-to-noise ratio between the spliced segments is close or even the same, the performance of the noise levels based audio splicing detection methods will decrease sharply. In addition, based on the fact that inserting an audio segment into another audio recording leads to anomalous variations of the electric network frequency (ENF) signal, several kinds of research [7][8][9] have shown that it is an efficient way to detect spliced audio via the analysis of ENF signal. Whereas due to legal restrictions, it is difficult to obtain concurrent reference datasets of power systems, which makes the ENF based audio splicing detection methods difficult to implement [10].…”