Estimation and Prevention of Sensor Replacement Attacks in Supervisory Control Systems

Abstract

This article proposes a defense method against an attacker who can replace the readings of vulnerable sensor events. We assume that the attacker can make a prediction action of the attack result, called an attack estimation, during the reconnaissance phase. Two concepts, strong sensor replacement attack estimability and weak SR-estimability, are introduced to capture the cases that the attacker can accurately and inaccurately estimate that an unsafe state is reached in an attacked system, respectively. A structure called an attacker estimator is built to model the behavior of a system under all stealthy attack decisions, from which two methods are derived to verify the proposed concepts. Based on honeypot technology, we construct a prevention module that can use decision insertion functions (DI-functions) to tamper with an attacker’s attack estimation and maintain the behavior of the closed-loop system. We design a two-player game-like structure (decision insertion structure) to describe the interaction between the module and the system. An algorithm is proposed to synthesize DI-functions for two different objectives.