This paper will consider the use of nonlinear dynamic forecasting techniques to extract message signals from a six-dimensional, message-modulated hyperchaotic communication scheme. It will be shown that a variety of messages can be successfully extracted using only a three-dimensional reconstruction. Further, the robustness of the signal extraction is demonstrated by extracting a speech message with the speech amplitude varied by over an order of magnitude. In all cases intelligible speech was extracted, so security was not greatly improved by going to a hyperchaotic system or larger amplitude signals.

• Received 22 January 1998

DOI:https://doi.org/10.1103/PhysRevE.58.1159