Gaussian mixture modeling for detecting integrity attacks in smart grids

Abstract

The thematics focusing on inserting intelligence in cyber-physical critical infrastructures (CI) have been receiving a lot of attention in the recent years. This paper presents a methodology able to differentiate between the normal state of a system composed of interdependent infrastructures and states that appear to be normal but the system (or parts of it) has been compromised. The system under attack seems to operate properly since the associated measurements are simply a variation of the normal ones created by the attacker, and intended to mislead the operator while the consequences may be of catastrophic nature. Here, we propose a holistic modeling scheme based on Gaussian mixture models estimating the probability density function of the parameters coming from linear time invariant (LTI) models. LTI models are approximating the relationships between the datastreams coming from the CI. The experimental platform includes a power grid simulator of the IEEE 30 bus model controlled by a cyber network platform. Subsequently, we implemented a wide range of integrity attacks (replay, ramp, pulse, scaling, and random) with different intensity levels. An extensive experimental campaign was designed and we report satisfying detection results.