Data-Driven False Data Injection Attacks Against Power Grids: A Random Matrix Approach

We address the problem of constructing false data injection (FDI) attacks that can bypass bad detector (BDD) a power grid. The attacker is assumed to have access only grid measurement traces collected over limited period time and no other prior knowledge about Existing related algorithms are formulated under assumption has measurements long (asymptotically infinite) period, which may not be realistic. show these approaches do perform well when from window only. design an enhanced algorithm construct FDI attack vectors in face nevertheless BDD with high probability. guided by results random matrix theory. Furthermore, we characterize important trade-off between attack's BDD-bypass probability its sparsity, affects spatial extent must achieved. Extensive simulations using MATPOWER simulator benchmark IEEE bus systems validate our findings.