Skip to main content

Table 2 Attack types and specifications

From: Effective blind speech watermarking via adaptive mean modulation and package synchronization in DWT domain

Item Type Description
A Resampling Conducting downsampling to 8 kHz and then upsampling back to 16 kHz.
B Requantization Quantizing the watermarked signal to 8 bits/sample and then back to 16 bits/sample.
C Low-pass filtering (I) Applying a low-pass filter with a cutoff frequency of 4 kHz.
D Low-pass filtering (II) Applying a low-pass filter with a cutoff frequency of 1 kHz.
E High-pass filtering Applying a high-pass filter with a cutoff frequency of 1 kHz.
F Amplitude scaling Scaling the amplitude of the watermarked signal by 0.85.
G Noise corruption (I) Adding zero-mean white Gaussian noise to the watermarked audio signal with SNR = 30 dB.
H Noise corruption (II) Adding zero-mean white Gaussian noise to the watermarked audio signal with SNR = 20 dB.
I DA/AD conversion Converting the digital audio file to an analog signal and then resampling the analog signal at 16 kHz. The DA/AD conversion is performed through an onboard Realtek ALC892 audio codec, of which the line-out is linked with the line-in using a cable line during playback and recording.
J Echo addition Adding an echo signal with a delay of 50 ms and a decay to 5% to the watermarked audio signal.
K Jittering Randomly deleting or adding one sample for every 100 samples within each frame.
L G.722 speech coding Encoding and decoding the watermarked speech signal with a G.722 wideband audio codec at 64 kbps.
M G.726 speech coding Encoding and decoding the watermarked speech signal with a G.726 wideband audio codec at 32 kbps.
N Time-shift by 1 sample Purposely shifting the watermarked audio signal by one sample.