Dynamic Reduced-Order Observer-Based Detection of False Data Injection Attacks With Application to Smart Grid Systems

Deep Latent Space Clustering for Detection of Stealthy False Data Injection Attacks Against AC State Estimation in Power Systems

“…The authors of [1] identified for the first time that stealthy FDI attack vectors, that can circumvent the conventional BDD algorithms, can be generated from the Jacobian matrix of the DC state estimation algorithms for power system. After that, numerous works have been proposed for the detection of FDI attacks using graph theoretic models [2], [3], sparse optimization [4], Kalman Filters [5]- [7] and a variety of different algorithms, each offering some unique advantages. Recently an internal observer based detection algorithm was proposed in [8] for FDIA attacks.…”

“…The authors in [6] further use Kalman Filters to estimate unknown time varying attack parameters and for the recovery of state estimates. In a different approach, authors in [3] develop a bank of dynamic reduced-order observers by leveraging graph theoretic properties of an interconnected smart grid for attack detection by generating residual signals. The power system is modelled as a discrete time linear dynamical system in [7].…”

Decentralized attack detection for multi‐area power systems via interconnection‐decoupled sliding mode observer

“…The attack detection problem has been considered for various cyber‐attacks on CPSs as in References 8‐11, in which the false data injection attacks (FDIAs) have received increasing attention due to their widely varied types and catastrophic impact 12,13 . As pointed out in Reference 14, existing attack detection methods for revealing FDIAs on CPSs can be roughly classified into data‐driven and knowledge‐based schemes.…”

“…Considering the complicated and large‐scale characteristics of power systems, the control design (e.g., load frequency control (LFC)), state estimation, and economic dispatching are expected to be achieved in a decentralized or distributed way. Based on this fact, decentralized and distributed detection schemes have been recently investigated as in References 10,26‐28. For instance, the authors in Reference 26 proposed a state and attack observer for LFC system to realize decentralized attack detection and mitigation, noting that the load variation was assumed to be known, which makes it less applicable.…”

“…For instance, the authors in Reference 26 proposed a state and attack observer for LFC system to realize decentralized attack detection and mitigation, noting that the load variation was assumed to be known, which makes it less applicable. In Reference 10, an $$$ {L}_{\infty } $$$ dynamic reduced‐order observer‐based decentralized detection scheme was adopted to reduce the conservativeness of the residual threshold and improve the detectability to FDIAs. To detect more complex cyber‐attacks, two distributed attack detection methods were investigated in References 27 and 28, respectively, in which the detector contained a decentralized unknown input observer (UIO) and a distributed Luenberger‐like observer.…”