Quality-driven secure audio transmissions in wireless multimedia sensor networks

Many audio applications such as audio surveillance and human acoustic health monitoring require security protections for audio streaming over WSNs. The process of watermarking which embeds small amounts of data (i.e., the watermark) into the original audio is an effective technique to ensure the integrity of received audio data at the receiver in energy-constrained WSNs. However, the selection of positions to embed watermark into audio streams is critical to both received audio quality and watermarking authentication performance in error-prone wireless transmission environments. In this paper we propose an approach that dynamically determines the range of middle sub-band components for embedding the watermark with minimum quality distortions, based on psycho-acoustic models and adaptive sub-band thresholds. In addition, through unequal network resource allocation schemes the proposed approach protects both middle sub-bands and high sub-bands, which include the important audio components. Our theoretical analysis and simulation results demonstrate that the proposed quality-driven energy-efficient watermarking approach for audio transmissions can achieve considerable performance gains in WSNs.