Globally Stealthy Attacks Against Distributed State Estimation in Smart Grid

With the continuous expansion of grid nodes, traditional centralized methods exhibit certain limitations in the amount of communication data and computational cost for state estimation. In the past few decades, distributed state estimation has been fully developed in multi-area smart grid (SG). According to different areas or electrical equipment, the SG is divided into several nodes, and each node can estimate the entire state of the SG through local sensor measurement information and neighbor state estimation information. Although malicious attackers can use the different information between local sensor data and neighbor state estimation communication data to attack grid nodes, it will inevitably cause the attack detectors of other nodes to trigger alarms due to the consistency of grid state estimation. In this paper, we propose a globally stealthy attack strategy that biases the state estimation of the attacked node without triggering alarms on all nodes’ detectors. We provide sufficient and necessary conditions to implement a globally stealthy attack on grid nodes and extend it to multi-node attacks in order to reduce the number of communication links and increase the degree of damage. We also give the corresponding attack algorithm to realize these attacks. Finally, our proposed globally stealthy attack methods are simulated and verified in IEEE 39-bus power system.
Note to Practitioners
—In this paper, we propose two attack strategies against distributed state estimation in smart grids and present the corresponding attack methods. Since the grid is a multi-area power system, attacks between different nodes are more likely than attacks within nodes. Most of the previous literature focused on the ability of an attack to remain stealthy within the attacked node. However, due to the presence of the consistency term of the distributed state estimation, the state of the attacked node will be propagated to other nodes and detected by other nodes. Therefore, in this paper, we consider an approach that can maintain stealthy at all nodes. The effectiveness of the proposed attack strategy against distributed state estimation in smart grid is demonstrated in IEEE 39-bus power system. It is shown that a globally stealthy attack leads to a bias in the state estimation of the attacked node and does not trigger an alarm at any node. In summary, the proposed attack strategy is useful for the research community to design targeted detection strategies and can be conveniently applied to real-world security management systems in smart grids.