Search papers, labs, and topics across Lattice.
This paper introduces AdaptoNet, a modular neural network designed to enhance cyber-physical attack detection in power grids by adapting to varying measurement availability. The authors highlight that traditional data-driven methods suffer a drastic performance drop of over 86% during data denial attacks, which target specific regions while causing line outages. Through a combination of a frozen foundational module and a trainable adaptive module, AdaptoNet significantly improves detection performance, achieving an F1 score recovery from below 12% to over 81% under challenging conditions, thus approaching the baseline performance of 89%-99% with complete data.
AdaptoNet recovers detection performance from below 12% to over 81% in the face of targeted cyber attacks, showcasing a revolutionary approach to resilience in power grid security.
Cyber attacks on the power grid combine physical disruptions with compromised data to destabilize cyber-physical systems. We demonstrate that data denial attacks, where adversaries block measurements in a targeted region while triggering a line outage, reduce detection performance by more than 86\%, rendering standard data-driven methods ineffective. We propose AdaptoNet, a modular neural network that adapts to measurement availability through conditional controls. AdaptoNet pairs a frozen foundational module trained on complete data with a trainable adaptive module, conditioned on a binary measurement-availability vector, enabling the model to distinguish between denied and anomalous data without retraining the foundational module. Evaluated across four IEEE test systems (30-, 39-, 57-, and 118-bus) under in-region attacks blocking up to 20% of measurements, AdaptoNet recovers F1 from below 12\% to above 81\%, an approximate sevenfold improvement approaching the 89%-99% baseline with complete measurements.
