Security Metrics of Networked Control Systems^footnoteinfo
(extended preprint)

Recently, there has been significant interest and work in the broad area of security of Networked Control Systems (NCSs), see, e.g., [1, 2, 7, 21, 23, 26, 28, 29, 30, 32, 38]. This topic investigates properties of conventional control systems in the presence of adversarial disturbances. Control theory has shown great ability to robustly deal with disturbances and uncertainties. However, adversarial attacks raise all-new issues due to the aggressive and strategic nature of the disturbances that attackers might inject into the system.
This paper focuses on quantifying and minimizing attacker capabilities in NCSs. A majority of the work on attack detection leverages the established literature of fault detection [7, 9, 22, 32]. A fault detection approach uses an estimator to forecast the evolution of the system dynamics. When the residual (the difference between measurements and their estimates) is larger than a predetermined threshold, an alarm is raised. Arguably the most insidious attacks are those that occur without our knowledge. Fault detectors impose limits on attacks if the attacker aims at avoiding being identified. Beyond retooling these existing methods for the new attack detection context, a fundamental question is: given a chosen fault detection approach, how does this method constrain the influence of an attacker? More specifically, what is an attacker able to accomplish when a system employs certain fault detection procedure?
Different methodologies exist for evaluating the impact of attacks. Most of the existing work uses some measure of state deviation. A number of groups have studied the system response when attacks are constrained by the detector, i.e., they investigate the system trajectories that can be induced due to stealthy attacks – attacks such that the detector threshold is never crossed [8, 12, 14, 16, 28, 31, 32]. In this manuscript, for given system dynamics, we provide mathematical tools for quantifying and minimizing the potential impact of sensor stealthy attacks on the system dynamics. We consider the set of states that stealthy attacks can induce in the system (the attacker’s stealthy reachable set) and use the “size” of this set as a security metric for the NCS. Stealthy reachable sets provide a metric of the system performance degradation induced by stealthy attacks. Because it is not mathematically tractable to compute these sets exactly, we provide analysis tools – in terms of Linear Matrix Inequalities (LMIs) – for computing ellipsoidal outer approximations of the attacker’s reachable sets. The obtained approximations quantify the attacker’s potential impact when it is constrained to stay hidden from the detector. We use the size (in terms of volume) of these ellipsoidal approximations to approximate the proposed security metric. As a second security metric, we propose the minimum distance from the attacker’s reachable set to a possible set of critical states – states that, if reached, compromise the integrity or safe operation of the system. We approximate this distance by the minimum distance between the ellipsoidal approximations and the critical states. This distance gives us intuition on how far the actual attacker’s reachable set is from the critical states. Once we have provided a complete set of analysis tools to approximate the aforementioned security metrics, we use these tools to derive synthesis tools (in terms of semidefinite programs) to redesign controllers and fault detectors such that the impact of stealthy attacks is minimized and the required attack-free system performance is guaranteed.
There are a few results in this direction already; chiefly the work in [27] (and the preliminary paper [26]), where the authors provide a recursive algorithm to compute ellipsoidal approximations of attacker’s reachable sets for Linear Time Invariant (LTI) systems subjected to Gaussian noise. The authors in [27] give analysis-only results for a very particular structure of controllers and fault-detectors. They consider Kalman-filter based fault detectors and use the state of the filter to construct output feedback controllers. Although this results in compact designs of controllers and fault detectors, the flexibility of having dedicated controllers and detectors (mainly for synthesis of secure control systems) is limited. We remark that, in the stochastic setting considered in [27], the detector threshold is always crossed even when there are no attacks. This is due to the infinite support of the Gaussian noise they consider. Thus, they do not consider stealthy attacks in the sense described above. Instead, they consider attacks that increase the alarm rate of the detector by a small amount only. Then, they approximate the attacker’s reachable set corresponding to this small increase.
The main contributions of this manuscript (in contrast to the work in [27]) are the following: 1) we provide a set of mathematical tools in terms of semidefinite programs to approximate reachable sets induced by stealthy attacks for LTI systems driven by peak bounded deterministic perturbations; 2) we provide both analysis and synthesis results for dedicated general dynamic output feedback controllers and observer-based fault detectors; 3) we propose two security metrics to assess the vulnerability of systems to attacks, and optimize these metrics (enhancing thus the system resilience to attacks) by synthesizing optimal controllers and detectors; 4) the synthesis part considers the attack-free performance of the closed-loop dynamics, i.e., we optimize the security metrics subject to certain prescribed attack-free system performance. In our preliminary work [31], we also approximate reachable sets of false-data-injection attacks but we consider the same stochastic framework as the one proposed in [27], i.e., Gaussian noise, joint Kalman-filter based fault detectors and controllers, and attacks increasing the alarm rate of the detector. Thus, the problems considered in this manuscript (and the obtained results) and the ones addressed in [31] are fundamentally different; and the set of results (and the tools used to obtain them) are different too. Moreover, in [31], we consider attacks to all the sensors. Although the latter case provides a worse-case scenario, we lose the capability of quantifying the sensitivity of the system dynamics to attacks on specific sensors. As in [27], the results in [31] mainly focus on analysis (although they hint how to address synthesis for joint Kalman-filter based detectors and controllers).
There are a few other results that considers different security metrics for control systems. All of them are fundamentally different to the work presented here. For instance, in [2, 3], for arbitrary detection procedures, the authors quantify how much the attacker can increase the asymptotic covariance (their security metric) of state estimates while remaining stealthy. They characterize stealthiness using the Kullback-Leibler Divergence [33] between the attack-free and the attacked estimates. In [37, 39], the authors use the notion of security index for LTI systems. This index refers to the smallest number of sensors and actuators that have to be compromised for successfully launching stealthy attacks. For linear stochastic systems, the authors is [24] propose two security metrics: the probability that some of the critical states leave a safety region; and the expected value of the infinity norm of the critical states. Finally, in [25], tools from finance risk theory are used to quantify security of LTI systems.
The remainder of the paper is organized as follows. In Section 2, we present some preliminaries results needed for the subsequent sections. We provide tools for computing outer time-varying bounds on the trajectories of a class of perturbed nonlinear discrete-time systems. Then, we use these tools to obtain outer ellipsoidal approximations of reachable sets of LTI systems driven by multiple peak bounded perturbations. The system dynamics, monitor, and controller descriptions are given in Section 3. Our proposed security metrics and analysis tools, together with some numerical results, are given in Section 4; and the corresponding synthesis results are given in Section 5. Finally, conclusions and recommendations are stated in Section 6.

In this section, we present some preliminary results needed for the subsequent sections. First, in Lemma 1, we present a preliminary tool used to compute outer time-varying bounds on the trajectories of perturbed discrete-time systems. Next, in Proposition 1, we use this lemma to compute outer ellipsoidal approximations of reachable sets of LTI systems driven by multiple peak bounded perturbations.

Proposition 1 provides a tool for computing time-varying ellipsoidal outer approximations ${\tilde{\mathcal{E}}}^{\xi}_{k}$ of $\mathcal{R}^{\xi}_{k}$. Note that ${\tilde{\mathcal{E}}}^{\xi}_{k}$ could be an arbitrarily conservative approximation of $\mathcal{R}^{\xi}_{k}$ as long as $\mathcal{R}^{\xi}_{k}\subseteq\tilde{\mathcal{E}}^{\xi}_{k}$. Then, to make ${\tilde{\mathcal{E}}}^{\xi}_{k}$ less conservative, we aim at obtaining ellipsoids with minimal volume, i.e., the tightest possible ellipsoid bounding $\mathcal{R}^{\xi}_{k}$ among all the ellipsoids generated by Proposition 1. To find such an ellipsoid, we look to minimize $(\det[\mathcal{P}])^{-1/2}$ subject to (6) because $(\det[\mathcal{P}])^{-1/2}$ is proportional to the volume of the asymptotic ellipsoid $\xi^{T}\mathcal{P}\xi=(N-a)/(1-a)$ for any $N\in{\mathbb{N}}$ and $a\in(0,1)$ [5]. We minimize $\log\det[\mathcal{P}^{-1}]$ instead as it shares the same minimizer with $(\det[\mathcal{P}])^{-1/2}$ and because for positive definite $\mathcal{P}$ this objective is convex [5]. This is stated in the following corollary of Proposition 1.

In Figure 1, we illustrate the potential tightness of the ellipsoidal outer approximations obtained using Corollary 1. The solid gray area is the actual reachable set obtained by extensive Monte Carlo simulations, and the ellipsoidal approximation is depicted with dashed lines. This figure corresponds to an LTI system driven by two peak bounded perturbation. The exact numerical values of the system matrices and the perturbations’ bounds can be found in [16, Section 4].

In this section, we introduce the system dynamics that we consider, the monitor that we use to pinpoint attacks, and the control scheme.

In this section, we provide tools for quantifying (for given $(L,A^{c},B^{c},C^{c},D^{c})$) and minimizing (by redesigning $(L,A^{c},B^{c},C^{c},D^{c})$) the impact of the attack $\delta_{k}$ on the state of the system when the monitor (16) is used for attack detection. We are interested in attacks that keep the monitor from raising alarms. This class of attacks is what we refer to as stealthy attacks. Here, we characterize ellipsoidal bounds on the set of states that stealthy attacks can induce in the system. In particular, we provide tools based on Linear Matrix Inequalities (LMIs) for computing ellipsoidal bounds on the reachable set of the attack sequence given the system dynamics, the control strategy, the system monitor, and the set of sensors being attacked.

The attacker can compromise up to $s$ sensors, $s\in\{1,\ldots,m\}$, of the system. Consider the monitor (16) and write $z_{k}$ in terms of the estimation error $e_{k}$ and $\delta_{k}$:

where $\Pi^{\frac{1}{2}}$ is the symmetric square root matrix of $\Pi$ and $\left\lVert\cdot\right\rVert$ denotes Euclidian norm. The set of feasible attack sequences that the attacker can launch while satisfying $z_{k}\leq 1$ (i.e., without raising alarms by the monitor) can be written as the constrained control problem on $\delta_{k}$:

Define the extended state $\zeta_{k}:=((x_{k}^{p})^{T},(x_{k}^{c})^{T},e_{k}^{T})^{T}$ and denote by $\psi^{\zeta}_{\delta}(k,\zeta_{k^{*}},\eta(\cdot),v(\cdot),\delta(\cdot))$ the solution of (18) at time instant $k\geq k^{*}$ given the extended state at the starting attack instant $\zeta_{k^{*}}$ and the infinite disturbance and attack sequences $\eta(\cdot):=\{\eta_{1},\eta_{2},\ldots\}$, $v(\cdot):=\{v_{1},v_{2},\ldots\}$, and $\delta(\cdot):=\{\delta_{1},\delta_{2},\ldots\}$. Let $\psi^{x}_{\delta}(k,\zeta_{k^{*}},\eta(\cdot),v(\cdot),\delta(\cdot))$ be the partition of $\psi^{\zeta}_{\delta}(k,\zeta_{k^{*}},\eta(\cdot),v(\cdot),\delta(\cdot))$ corresponding to the plant trajectories, i.e., the solution $x^{p}_{k}$ of (18). We are interested in the state trajectories that the attacker can induce in the system restricted to satisfy (20). To this end, we introduce the notion of stealthy reachable set:

In this manuscript, we propose to use the volume of the set $\mathcal{R}_{\Gamma,k}^{x}$ as a security metric. However, in general, it is not tractable to compute $\mathcal{R}_{\Gamma,k}^{x}$ exactly. Instead, we look for an outer approximation $\mathcal{E}^{x}_{\Gamma,k}$ satisfying $\mathcal{R}^{x}_{\Gamma,k}\subseteq\mathcal{E}^{x}_{\Gamma,k}$ for all $k\geq k^{*}$. In particular, for some positive definite $\mathcal{P}_{\Gamma}^{x}\in{\mathbb{R}}^{n\times n}$ and nonnegative function $\alpha_{k}^{x}$, we look for outer ellipsoidal approximations of the form $\mathcal{E}_{\Gamma,k}^{x}=\{x^{p}\in{\mathbb{R}}^{n}|(x^{p})^{T}\mathcal{P}_{\Gamma}^{x}x^{p}\leq\alpha_{k}^{x}\}$ such that $\mathcal{R}_{\Gamma,k}^{x}\subseteq\mathcal{E}_{\Gamma,k}^{x}$. That is, the ellipsoid $(x^{p})^{T}\mathcal{P}_{\Gamma}^{x}x^{p}=\alpha_{k}^{x}$ contains all the possible trajectories that stealthy attacks of the form (20) can induce in the system. Because, for LTI systems $\mathcal{E}_{\Gamma,k}^{x}$ is a good approximation of $\mathcal{R}_{\Gamma,k}^{x}$, and because $\mathcal{E}_{\Gamma,k}^{x}$ can be computed efficiently using LMIs, we use the volume of $\mathcal{E}_{\Gamma,k}^{x}$ as an approximation of the proposed security metric. This approximation allows us to quantify the potential “damage” that sensor attacks can induce to the system in terms of the set of sensors being compromised (the attacker’s sensor selection matrix $\Gamma$). In Figure 3, we depict a schematic representation of the proposed ideas.