We study aperiodic stochastic resonant data storage in an extended system evolving on directed small-world
networks. Each node of the network represents a dynamical bistable system, and nodes are randomly connected
by the directed shortcuts with a rewiring probability. The constructive role of the internal noise and the random
connectivity is characterized by the bit error rate and demonstrated in numerical simulations. Random internal
noise in each node enhances the survival of a short-time length of binary signal via aperiodic stochastic resonance.
Interestingly, random connectivity further improves the propagation time of binary information through the
A DCT-domain watermarking scheme, based on nonlinear bistable detectors, is presented. A binary copyright character, i.e. watermark, is firstly reordered into a binary zig-zag sequence, and then mapped into the pulse amplitude modulated waveforms. Certain desyn-chronization time can be arbitrarily placed into one code of the
modulated signal, and will be tolerated due to the robust superiority of nonlinear detectors over matched filters. The watermarking signal is then embedded in a selected set of DCT coeficients of an image in medium frequency domain. The selected set of DCT coeffcients is shuffled by Arnold transform and looks more like background
noise for the watermark signal. The copyright character can be extracted by the nonlinear bistable detector without resorting to the original image, i.e. blind watermark detection. Interestingly, more higher similarity between the original character and the extracted one can be further achieved by a parallel array of bistable
detectors via the mechanism of array stochastic resonance. Efficacy of the proposed watermarking scheme is proved on some common attacks in experiments.