Design of Data-Injection Attacks for Cyber-Physical Systems Based on Kullback–Leibler Divergence

Hongyang Li,Junfeng Zhang,Xiao He
DOI: https://doi.org/10.1016/j.neucom.2019.05.085
IF: 6
2019-01-01
Neurocomputing
Abstract:In this paper, the design problem of data-injection attacks is investigated for a class of cyber-physical systems (CPSs) whose inputs can be changed by the malicious adversary. A novel attack design method is proposed by solving an optimization problem based on Kullback–Leibler divergence (KLD). It can make maximum influence on the residual sequences in the sense of probability distribution but cannot be detected by the detector with preset significance level. A simulation study is carried out to demonstrate the effectiveness and applicability of the proposed method.
What problem does this paper attempt to address?