Neutralizing zero dynamics attack on sampled-data systems via generalized holds

Abstract

Zero dynamics attacks can be lethal to cyber–physical systems because they can be harmful to physical plants and impossible to detect. Fortunately, if the given continuous-time physical system is minimum phase, the attack is not so effective even if it cannot be detected. However, the situation can become unfavorable if one uses digital control by sampling the sensor measurement and using a zero-order hold for actuation because of the ‘sampling zeros.’ When the continuous-time system has a relative degree greater than two and the sampling period is small, the sampled-data system must have unstable zeros (even if the continuous-time system is minimum phase), so that the cyber–physical system becomes vulnerable to ‘sampling zero dynamics attack.’ In this paper, we present an idea to neutralize the zero dynamics attack for single-input and single-output sampled-data systems by shifting the unstable discrete-time zeros into stable ones. This idea is realized by employing the so-called ‘generalized hold’ which replaces a standard zero-order hold. It is shown that, under mild assumptions, a generalized hold exists which places the discrete-time zeros at desired positions. Furthermore, we formulate the design problem as an optimization problem whose performance index is related to the inter-sample behavior of the physical plant, and propose an optimal gain which alleviates the performance degradation caused by generalized hold as much as possible.