THESIS
2018
xviii, 108 pages : illustrations ; 30 cm
Abstract
This thesis investigates security issues for remote state estimation in the context
of cyber-physical systems. Different problems are studied from the perspective
of either a malicious attacker or a system designer. In the problem from the
attacker's perspective, a sensor monitors a dynamic process and transmits its
measurement to a remote estimator through a wireless network where a malicious
attacker may intercept and modify the transmitted data. A false-data
detector is adopted to monitor system behaviors and check data anomalies. We
propose an innovation-based integrity attack and present the feasibility constraint
which guarantees the attack stealthiness. Under the proposed attack,
the recursion of the remote estimation error covariance is derived. A closed-form
expressio...[
Read more ]
This thesis investigates security issues for remote state estimation in the context
of cyber-physical systems. Different problems are studied from the perspective
of either a malicious attacker or a system designer. In the problem from the
attacker's perspective, a sensor monitors a dynamic process and transmits its
measurement to a remote estimator through a wireless network where a malicious
attacker may intercept and modify the transmitted data. A false-data
detector is adopted to monitor system behaviors and check data anomalies. We
propose an innovation-based integrity attack and present the feasibility constraint
which guarantees the attack stealthiness. Under the proposed attack,
the recursion of the remote estimation error covariance is derived. A closed-form
expression of the optimal attack strategy that maximizes the remote estimation
error covariance is obtained. The problem is then generalized from the following
two aspects. On the one hand, we consider an ∈-stealthy innovation-based
integrity attack, where the Kullback-Leibler divergence is used as a stealthiness
metric. Based on the evolution of remote estimation error covariance, a
two-stage optimization problem is formulated to investigate the optimal attack
strategy. We first prove that the optimal attack is Gaussian distributed and
provide a closed-form expression of the corresponding covariance matrix. Then,
the optimal attack is obtained by semi-definite programming. On the other
hand, we consider an innovation-based integrity attack under different information
sets. It is assumed that the attacker is not only able to compromise the
transmitted data but also able to measure the system state itself. The attack
strategy thus can be designed based on the intercepted data, the sensing data,
or alternatively the combined information. For each attack scenario, we analyze
the remote estimation performance and obtain the optimal attack strategy. For
scalar systems, we further derive the closed-form expressions of the optimal attacks
and compare the attack consequences between different information sets.
The secure state estimation problem from the system's perspective is studied
in a multi-sensor system. Suppose there are N sensors measuring the same
dynamic process and a subset of the sensors can potentially be compromised
by an attacker. To locate the compromised sensors and obtain a robust state
estimate, we propose a Gaussian-mixture-model-based detection algorithm. It
is able to cluster the local state estimate autonomously and provide a belief for
each sensor, based on which measurements from different sensors can be fused
accordingly. The performance of the proposed detection algorithm is evaluated
by the remote estimation performance and the average belief. The applications
of the proposed detection algorithm to other attack scenarios are also discussed.
Post a Comment