This paper considers a method of coding the sensor outputs in order to detectstealthy false data injection attacks. An intelligent attacker can design asequence of data injection to sensors and actuators that pass the stateestimator and statistical fault detector, based on knowledge of the systemparameters. To stay undetected, the injected data should increase the stateestimation errors while keep the estimation residues small. We employ a codingmatrix to change the original sensor outputs to increase the estimationresidues under intelligent data injection attacks. This is a low cost methodcompared with encryption schemes over all sensor measurements in communicationnetworks. We show the conditions of a feasible coding matrix under theassumption that the attacker does not have knowledge of the exact codingmatrix. An algorithm is developed to compute a feasible coding matrix, and, weshow that in general, multiple feasible coding matrices exist. To defendagainst attackers who estimates the coding matrix via sensor and actuatormeasurements, time-varying coding matrices are designed according to thedetection requirements. A heuristic algorithm to decide the time length ofupdating a coding matrix is then proposed.
展开▼
