Two-fold Intelligent Approach for Successful FDI Attack on Power Systems State Estimation

Abstract

Recent research works have revealed that state estimators in power systems are susceptible to false data injection attacks (FDIA). Still, for an adversary, constructing a least effort attack vector is difficult and known to be L0-norm optimization problem. In this paper, two-fold intelligent approach is proposed to optimally construct the FDIA vector. First, the problem of selecting the vector components is formulated as a constrained nonlinear programming problem and is solved using Genetic Algorithm. Second, a Neural Network is trained to generate in real-time the vector amplitudes. The attack vector is optimally selected in terms of number of measurements to compromise, the set of measurements accessible be the adversary, and flexibility to successfully pass Bad Data Detection algorithm of the state estimator. The performance of the attack vectors is analyzed on the IEEE 14-bus system against AC state estimator for a range of various system loading conditions and considering two attack strategies.