Bespoke Mitigation Framework for False Data Injection Attack-Induced Contingency Events

Perpetrators of large-scale Stealth False Data Injection Attacks (FDIA), who desire actual and accurate controllability of target Cyber Physical Power System (CPPS), often utilize Machine Learning (ML)-based systems to ascertain the minimum/optimal Control Signal Energy Cost (CSEC opt ) and Controllability Gramian Matrix (CGM opt )/principal submatrices (a.k.a., CGSM opt ), which turn out to be directly related to the involved convex relaxation. Likewise, those charged with FDIA mitigation are seeking to solve the same problem to anticipate optimal Target Nodes/Links (TN opt /TL opt ) for Targeted Control (TC opt ). During an attack, wherein conditions are fluid, it is a contest at machine speed between the ML systems of the attackers and defenders. As a Robust Convex Relaxation (RCR) is central to the described scenario, this paper proposes an FDIA mitigation framework, which is based upon a sequence of transformations — Nonnegative Matrix Factorization (NMF) to Multiresolution Matrix Factorization (MMF) to Corresponding Wavelet Transform (CORWT) to an Enhanced CORWT (ECORWT) — and a bespoke Robust Constriction Factor Particle Swarm Optimization (RCF-PSO)-based RCR (RPR), which leverages translation-invariant Continuous Wavelet Transforms (CWTs) to achieve enhanced insight into the prospective CGM opt /CGSM opt .