An Efficient Tamper Detection and Recovery Scheme for Attacked Speech Signal

Abstract

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. A watermarking method is suggested to embed authentication codes, synchronization identifiers, and spare part data. 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, and the normalized correlation for the retrieved signal is close to attacked signals. Additionally, this approach achieves less distortion in the watermarked signal quality measured by log spectrum distortion, which is about 0.0072. The proposed approach is compared to the related methods, where it achieves the best results, including high robustness against different types of attacks