An Efficient Tamper Detection and Recovery Scheme for Attacked Speech Signal.

Data security is an essential communication issue, where the resistivity to data tampering, corruption, and attacks forms a critical problem. This paper proposes a robust approach for tamper detection in a speech signal and recovering the destroyed parts of the attacked signal from the embedded spare parts using a specific strategy. The given approach deals with the abnormalities in the quality of the message speech signal that occurs due to different attacks. To solve that issue, A watermarking method is suggested to embed self-extracted data of authentication codes, synchronization identifiers, and spare part data from the same speech signal. The authentication codes are generated using the Singular Value Decomposition (SVD) method, while G.723.1 speech CODEC generates the spare part data. The proposed strategy considers attacks such as muting and replacement. According to the experiments, the quality of the watermarked and retrieved signals has improved. The average signal-to-noise ratio is above 70 dB for the watermarked signal quality, and the normalized correlation for the retrieved signal is close to the original speech signal under different operations. The results state fully recovering rate for the attacked signal under different types and lengths of attacks (10%-80% of Tampering Rates). Additionally, this approach achieves less distortion in the watermarked signal and the recovered signal quality measured by log spectrum distortion, which is about 0.0072 and 0.225 (at high Tampering Rate), respectively. The proposed approach is compared to the related methods, where it achieves the best results, including high robustness against different types of attacks. [ABSTRACT FROM AUTHOR]