INTELLIGENT NETWORK COMPUTING AUGMENTED BY CHAOTIC NEURAL NETWORKS, BLIND SOURCES SEPARATION AND DIGITAL WATERMARKS

Abstract

A powerful engine of Information Acquisition (IA) that can automatically extract feature and discover knowledge from the real world environment is known as an Intelligent Information Acquisition, IIA. A typical class of such a system is called smart sensor web, e.g. a dynamical smart dust scattering over a field in a self-organization fashion, an electrical powerline serving the volume surveillance in the city public environment where the powerline plug output is the power source required for the operation of the sensors and the input to the powerline is the ecosystem-like communication sparse change signal without the address — a topic of previous paper in this journal. This is possible because IIA of SSW must satisfy 6 W knowing whom, where, when, how, why and what to do. In that sense it will also satisfy the new ad hoc Internet protocol which demands, besides security and authenticity, the privacy protection — a topic of this paper. Such an IIA can serve medical patients, bank customers, or any other Special Interest Groups (SIGs) who must communicate among themselves at the public channel but are not yet ready to share information with the public. Our design might ward off invasive search engine to protect the strict privacy which requires literally no involvement of any third party besides the sender and receiver. Thus, similar to classical postal system, we have adopted envelop wrapper, called digital watermark (DWM), to hide the address of SpatioTemporal Key (STK), so that we do not need to send the high-precision chaos initial value and Chaotic Neural Net (CNN) typing but only the binary "Address Initiation (AI)" listed inside a PC library routine shared by 2N members of SIGs (N>1). The address bits can be redundant in a distributed repetition in order to tolerate any channel noise, package loss, and JPEG-related image compression. Furthermore, the bits are hidden over almost infinitely long lexicographical text pixel sequence of video image frames as digital watermarks. For example, the simplest coded message is the output of "EXOR" operation between the computed STK and originally intended message. Especially, the self-inverse decoding enjoys the projection operator, a cycle two reduction EXOREXOR = I, that means the recovered message comes simply from an identical bits-parallel "EXOR" operation between received coded message and the receiver-on-site-derived STK. Our rigorous definition of privacy enjoys also reasonable security which can be measured in NSA exhausted search sense of a few days up to years of CPU time, and also the cost saving without relying on commercial security RSA codec.