Abstract

We present a set of watermarking techniques for effective covert communication over an audio signal carrier. Watermark robustness is enabled using: (i) redundant spread-spectrum for prevention against desynchronization attacks and (ii) psycho-acoustic frequency masking (PAFM). We show that PAFM impacts the balance of "ones" and "zeros" in the part of the spread-spectrum sequence used for watermark detection and propose a modified covariance test to compensate for that anomaly. The covert message is combined with the spread-spectrum sequence and additionally permuted in time to improve correlation convergence as well as the security of individual message bits. We have incorporated these techniques into a system capable of reliably detecting watermarks in audio that has been modified using a composition of attacks that degrade the original audio characteristics well beyond the limit of acceptable quality. Such attacks include: fluctuating scaling in time and frequency, editing, recompression, noise addition, resampling, normalization, and filtering.

Citations