Data Embedding, the Method of Types, and Universal Receivers

Southern Methodist University, Dallas TX

Investigators

Abstract

The field of information hiding includes steganography, where a message is concealed in another data stream, and watermarking, where ownership data is included in a digital object to be protected. A third field of information hiding is the field of data embedding, wherein additional information is incorporated in the transmitted data stream by using a key and distorting (slightly) the original object. The embedded information cannot be reconstructed without the key. We propose an entirely new approach to data embedding based upon the method of types and universal receiver design. In our approach, a new data sequence is embedded in the original data stream using the method of types, and the embedded data is extracted using a type-based universal receiver. The choice of type and rate for the embedded data is based upon an analysis of portions of the original data stream. The universal receiver learns the type from the received data alone, and hence, there is no side information as in previous data embedding techniques. The embedding process and the receiver are both data adaptive, so the original data stream can be reconstructed without error. We investigate the fundamental limits of this approach to data embedding by analyzing the characteristics of the original data stream that facilitate data embedding, by studying which techniques allow the highest data rates to be embedded, and by examining the tradeoffs between embedded data rate and errors in the embedded data. This approach to data embedding offers the possibility of expanding the delivered data rate of many existing telecommunications links, both wired and wireless, without changing allocated bandwidths.