Processing math: 63%
A journal of IEEE and CAA , publishes high-quality papers in English on original theoretical/experimental research and development in all areas of automation
K. K. Zhang, C. Keliris, T. Parisini, B. Jiang, and M. M. Polycarpou, “Passive attack detection for a class of stealthy intermittent integrity attacks,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 4, pp. 898–915, Apr. 2023. doi: 10.1109/JAS.2023.123177
Citation: K. K. Zhang, C. Keliris, T. Parisini, B. Jiang, and M. M. Polycarpou, “Passive attack detection for a class of stealthy intermittent integrity attacks,” IEEE/CAA J. Autom. Sinica, vol. 10, no. 4, pp. 898–915, Apr. 2023. doi: 10.1109/JAS.2023.123177

Passive Attack Detection for a Class of Stealthy Intermittent Integrity Attacks

doi: 10.1109/JAS.2023.123177
Funds:  This work was supported by the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skodowska-Curie (101027980 (CSP-CPS-A-ICA), 739551 (KIOS CoE-TEAMING)), the Italian Ministry for Research in the Framework of the 2017 Program for Research Projects of National Interest (PRIN) (2017YKXYXJ), the National Natural Science Foundation of China (61903188, 62073165, 62020106003), the Natural Science Foundation of Jiangsu Province (BK20190403), the 111 Project (B20007), and the Priority Academic Program Development of Jiangsu Higher Education Institutions
More Information
  • This paper proposes a passive methodology for detecting a class of stealthy intermittent integrity attacks in cyber-physical systems subject to process disturbances and measurement noise. A stealthy intermittent integrity attack strategy is first proposed by modifying a zero-dynamics attack model. The stealthiness of the generated attacks is rigorously investigated under the condition that the adversary does not know precisely the system state values. In order to help detect such attacks, a backward-in-time detection residual is proposed based on an equivalent quantity of the system state change, due to the attack, at a time prior to the attack occurrence time. A key characteristic of this residual is that its magnitude increases every time a new attack occurs. To estimate this unknown residual, an optimal fixed-point smoother is proposed by minimizing a piece-wise linear quadratic cost function with a set of specifically designed weighting matrices. The smoother design guarantees robustness with respect to process disturbances and measurement noise, and is also able to maintain sensitivity as time progresses to intermittent integrity attack by resetting the covariance matrix based on the weighting matrices. The adaptive threshold is designed based on the estimated backward-in-time residual, and the attack detectability analysis is rigorously investigated to characterize quantitatively the class of attacks that can be detected by the proposed methodology. Finally, a simulation example is used to demonstrate the effectiveness of the developed methodology.

     

  • CYBER-PHYSICAL systems (CPS) integrate control, computation and communication techniques with physical engineered control systems [1]. Due to the emergence of such a complex integration, more security vulnerabilities in CPS arise and more malicious cyber threats greatly endanger various key aspects of CPS operation. A series of cyber attack events, such as the Stuxnet worm attack on the Iranian nuclear facilities, the attack on the Ukrainian power distribution network, and the recent colonial oil pipeline attack in USA, have taken place in recent years (more details and examples can be found in [2]–[5]). Therefore, state-of-the-art cyber attack diagnostic technologies are required to safeguard the operation of CPS against possible malicious attacks.

    Integrity in computer science refers to the trustworthiness of data, whereas in the context of CPS, integrity attacks compromise the integrity of the transmitted data of the CPS [1]. Replay attacks [6], covert attacks [7], [8], zero-dynamics attacks [9] and false-data injection attacks [10] are the most commonly studied stealthy integrity attacks. Several survey papers provide overviews of integrity attacks from a system control perspective (see, e.g., [2], [9], [11] and [12]). Recently, some new types of integrity attacks have also been investigated. For instance, in order to achieve perfect stealthiness, a self-generated approach is developed in [13] for generating particular false data, and [14] considers a class of closed-loop nonlinear systems and develops a stealthy integrity attack formulation approach for such systems. In the aforementioned integrity attack studies, false data are injected into the CPS continuously, whereas the problem of intermittent data-injection is overseen. It is important to note that the intermittent mode of attack injection may greatly affect both the stealthiness of an attack event and the power energy consumed by an attacker.

    Intermittent attacks are characterized by piece-wise attack signals. Denial-of-service attacks (DOS) [15]–[18] are typical examples of intermittent attacks. The available power energy to the attacker is optimally managed by scheduling the DoS attack application time instants in [16], [17], whereas the stealthiness of the attack is not considered. In this paper, we consider integrity attacks performed in an intermittent mode; these are referred to as intermittent integrity attacks. Unlike the strategy for typical continuous integrity attacks, the strategy for generating intermittent integrity attacks includes two steps: 1) constructing a stealthy attack model, and 2) scheduling attack-activating and attack-pausing time instants. In terms of attack models, typical stealthy integrity attack models such as zero-dynamics attack models [14], [19], replay attack models [6] and covert attack models [20], [21], can also act as models for intermittent integrity attacks since they are undetectable by typical anomaly detectors. The scheduling of intermittent attacks (activation and pausing time instants) can improve their stealthiness through activating the attacks for a limited amount of time such that anomaly detectors do not have sufficient time to detect them. Using attacks with specially scheduled activating and pausing time instants can also save the adversary’s instantaneous power energy by avoiding attack signal divergence requirements. A counterexample to this is the zero-dynamics attack case, in which the divergence attack signals are used (e.g., [9]), and may consume large instantaneous power energy. Moreover, intermittent integrity attacks can overcome the attack defense strategy such as topology switching in the case of multi-agent systems [22]. Particularly, the adversary may pause the attack prior to the switch of the topology, and then resume and update the attack to maintain the stealthiness in the new topology.

    In the past decade, several methods for detecting stealthy integrity attacks have been proposed by the research community, which fell into two categories: active and passive detection methods. In active detection methods, such as the watermarking and moving-target approaches, either authentication signals are injected into the information flows of CPS or secret modules are embedded to CPS loops and series-connected to physical plants (see, e.g., [23]–[27]). On the other hand, in passive attack detection methods, only analytical redundancy approaches are used to detect integrity attacks, without using any authentication signals or secret modules [28]. However, typical passive anomaly detectors are not able to provide desirable attack detection performance. Note that attack detectors are a special type of anomaly detectors specifically designed for detecting attacks. For example, fault detection schemes in [29]–[33] may not be able to detect stealthy integrity attacks (such as [6]–[9] and [13]). Generally, the reason for this is that, in the presence of a stealthy integrity attack, the information resources (sensory measurements and control inputs) of analytical redundancy approaches remain either unchanged or slightly altered and therefore, analytical redundancy approaches that are inherently not sufficiently sensitive to such slight changes, are unable to detect stealthy integrity attacks. In [34], multiple filters are combined to formulate a type of analytical redundancy-based passive attack detector, which can detect various types of stealthy integrity attacks. However, such a detector requires additional physical communication channels, which may not be feasible or realistic. Traditional analytical redundancy-based passive anomaly detectors are enhanced in [35] and [36] for detecting stealthy integrity attacks by using a backward-in-time signal processor. In these studies, even a small change due to a stealthy integrity attack is amplified by a backward-in-time signal processor such that the amplified change becomes sufficiently “large” to be detected. However, stealthy intermittent integrity attacks are not considered in [34]–[36].

    The detection of stealthy intermittent integrity attacks remains an open problem and few research works have been published. For example, [37] combines an analytical redundancy-based passive detector and a set-theoretic detector for detecting intermittent integrity attacks. In this method, the transient overshoots of the analytical redundancy and the strict detection guarantees of the set-theoretic detector are integrated to detect promptly the intermittent attacks. However, the stealthiness of the attacks is not considered in [37], which prevents the detection method from applied effectively in the case of stealthy intermittent integrity attacks.

    This paper utilizes the backward-in-time approach in the context of intermittent integrity attacks, and proposes an analytical redundancy-based passive detection methodology for detecting a class of stealthy intermittent integrity attacks. Specifically, the contributions of this paper are summarized as follows:

    1) A stealthy intermittent integrity attack generation strategy is formulated, which does not require that the adversary has precise knowledge of the system states. A backward-in-time detection residual is formulated, which increases in magnitude each time a new attack occurs;

    2) An optimal fixed-point smoother with covariance matrix resetting is proposed to implement the aforementioned backward-in-time residual. Such a smoother guarantees robustness to both disturbances and noise, and can also reset the covariance matrix to maintain sensitivity to intermittent integrity attacks;

    3) The corresponding adaptive threshold is designed, and an attack detectability analysis is carried out to characterize quantitatively the class of detectable stealthy intermittent integrity attacks.

    In terms of the stealthy intermittent integrity attacks, compared to [22], the attack generation proposed in this paper addresses the practical issue that the adversary does not have precise knowledge of the system states. In addition, in contrast with [38] in which the pausing and resuming time instants of intermittent integrity attacks are scheduled for saving power energy, this paper focuses on designing the attack generation strategy such that the generated attacks are stealthy, regardless of the pausing and resuming time instants.

    Compared to the authors’ previous work [36], the intermittency feature of stealthy integrity attacks is considered in this paper. Moreover, the designed smoother in this paper introduces the covariance matrix resetting technique, which is shown to guarantee robustness to both disturbances and noise, and simultaneously guarantee sensitivity to stealthy intermittent integrity attacks.

    Consider a vector signal x(t):R+Rn. Then, x(t)=0 for t[t1,t2]R+ means that x(t)=0 identically for all t[t1,t2]; x(t)0 for t[t1,t2]R+ means that x(t)0 for at least one time instant t[t1,t2]. The notation || is used in this paper to represent the absolute value for scalars, and the 2-norm for vectors and matrices. For a set S, |S| represents the number of the elements in S. A vector x(t)L2[t1,t2] if t2t1xT(τ)x(τ)dτ is finite. For a signal x(t) in the finite time interval [t1,t2] and a given matrix R0 with proper dimensions, we define x(t)2R=t2t1xT(τ)Rx(τ)dτ. For a constant vector x and a given matrix R0 with proper dimensions, we define x2R=xTRx. For a matrix A, ˉσ(A) and σ_(A) represent the maximum and minimum singular values of A respectively. Italics in the paper are used to highlight important associated sentences and terminologies.

    The rest of this paper is organized as follows. In Section II, the problem is formulated. Section III analyzes the stealthiness of the intermittent integrity attacks formulated in this work. In Section IV, the backward-in-time detection residual is introduced, and its theoretical feasibility to indicate the stealthy intermittent integrity attacks is presented. The design details of the implementable backward-in-time detection methodology are presented in Section V and the attack detectability analysis is shown in Section VI. Section VII presents a simulation example and finally, the conclusions are drawn in Section VIII.

    A general structure of a CPS subject to integrity cyber attacks is depicted in Fig. 1. It consists of a physical plant P, a feedback controller C, an anomaly detector D, an actuator communication network Na and a sensor communication network Ns.

    Figure  1.  Schematic diagram of CPS in the presence of integrity cyber attacks.

    In order to simplify the notation, the closed-loop CPS including C, P, Na and Ns are jointly denoted by W throughout the paper. The closed-loop system W is described by

    W:{˙x(t)=Ax(t)+Bu(t)+BΓuau(t)+Dω(t)u(t)=Ky(t)+yref(t)y(t)=Cx(t)+v(t)+Γyay(t) (1)

    where xRnx is the state vector, uRnu is the vector of control data generated by the controller C, yRny is the vector of sensor measurements received by the controller and the anomaly detector. The integers nx and ny satisfy ny<nx. Moreover, yref(t)Rny denotes the output reference signals, and ω(t)Rnω and v(t)Rny represent the vectors of process disturbances and measurement noise, respectively. Let Ku{1,,nu} and Ky{1,,ny} represent the disruption resources available to the attacker, i.e., the sets of actuator and sensor communication channels respectively that can be affected by the attacker. The distribution matrices ΓuBnu×|Ku| and ΓyBny×|Ky| (B={0,1}) are the binary incidence matrices mapping the attack signal to the respective channels. The attack signals are au(t)=[au,1(t),,au,|Ku|(t)]TR|Ku| and ay(t)=[ay,1(t),,ay,|Ky|(t)]TR|Ky|. For each i{1,,|Ku|}, au,i(t)=0 for tR+ if no attack occurs on the i-th transmission channel of Na, and similarly, for each j{1,,|Ky|}, ay,j(t)=0 for tR+ if the j-th transmission channel of Ns is not under attack. Throughout this paper, we denote jointly the attack signals au and ay as a(t)=[aTu(t),aTy(t)]T. We consider that the attack starts at an unknown time t=T0 and hence a(t)=0 for 0t<T0. By letting Ba=[BΓu,0nx×|Ky|] and Da=[0ny×|Ku|,Γy], then BΓuau(t)=Baa(t) and Γyay(t)=Daa(t).

    In addition, the matrices ARnx×nx, BRnx×nu, CRny×nx and DRnx×nω are system matrices known by the defender. The pair (A,D) is assumed to be stabilizable and (C,A) is assumed to be observable. The control gain KRnu×nx is designed to stabilize the system. In the nominal case (a(t)=0 for tR+), the closed-loop system Wn is given by

    Wn:{˙xn(t)=Axn(t)+Bun(t)+Dω(t)un(t)=Kyn(t)+yref(t)yn(t)=Cxn(t)+v(t) (2)

    where xn, un and yn represent the state, the control input and the output, respectively, in the nominal case.

    We consider that the CPS is equipped with a typical anomaly detector D (see Fig. 1) for detecting some normal anomalies (faults or attacks). Specifically, the anomaly detector D contains a detection residual r(t) and a constant threshold Jth. Without loss of generality, based on [39], we consider that the residual has the following form:

    r(t)=D(u(t),y(t),yref(t))y(t) (3)

    where D:Rnu×Rny×RnyRny is an observer (robust Luenberger observer [39], Kalman filter [29], adaptive observer [40] or sliding mode observer [41]) designed based on the analytical redundancy concept to generate an estimate of y(t) in the nominal case. By using the evaluation function J(t)=|r(t)|, the occurrence of an anomaly (fault or attack) is ascertained if at some time t>T0, J(t) exceeds the threshold Jth, i.e.,

    J(t)>Jth,alarmtriggering. (4)

    It should be noted that the residual r(t) contributing to indicate the occurrence of the anomaly is based on the estimates of the system outputs at the time posterior to the anomaly occurrence time. Hence, D is referred to as a forward-in-time detector in this paper. Most fault detectors in the literature such as [29], [30], [39], belong to the class of forward-in-time detectors.

    Throughout this work, to simplify the presentation, stealthiness is used to refer to stealthiness with respect to the typical standard anomaly detector D characterized by the residual r in (3) and the threshold Jth in (4). The attacker has the following available resources: partial model knowledge (A, BΓu, C) and disruption resources (Ky, Ku), where BΓu represents the partial columns of B corresponding to Ku. The strategy for generating stealthy intermittent integrity attacks includes two steps: 1) constructing attack models, and 2) scheduling attack-activating and attack-pausing time instants. For Step 1), the model used for generating stealthy integrity attack is given in Section III. For Step 2), in this paper, we consider that the activating and pausing time instants have been scheduled by the attacker, and the attack model is activated at the following time instants:

    t1,,tNa,NaN+.

    In addition, we consider that the k-th attack is active for a time length τk (dwell time) where 0<τktk+1tk and hence, it is inactive (attack silence) for ttk+τk. Then, the activating time interval Ωack (attack active) and the silence time interval Ωsik (attack silence) for the k-th attack can be given respectively as follows:

    Ωack=[tk,tk+τk),Ωsik=[tk+τk,+). (5)

    Also, we define an action time interval Ω0k that the k-th attack affects the system as follows:

    Ω0k=ΩackΩsik=[tk,+). (6)

    In addition, an auxiliary time interval “attack slot Ωk” is defined as follows:

    Ωk=[tk,tk+1),k{1,,Na1},ΩNa=[tNa,+). (7)

    A schematic of Ωk, Ωack, Ωsik and Ω0k is given in Fig. 2.

    Figure  2.  Schematic diagram of Ωk, Ωack, Ωsik and Ω0k.

    The first aim is to propose an intermittent attack model and then analyze rigorously the stealthiness of the generated intermittent integrity attacks by the model with respect to the typical anomaly detector D. The second objective of this paper is to design an attack detection methodology, based on a passive analytical redundancy approach (i.e., only using the system output y and the control input u in the system W). Such a methodology is able to overcome the stealthiness of the intermittent integrity attacks and thus, can detect their occurrences.

    This section proposes an intermittent integrity attack model and analyzes the stealthiness (with respect to the typical anomaly detector D) of the generated attacks. In this paper, the zero-dynamics attack model in [19] is modified to perform intermittent integrity attacks. Given the attack activating time interval Ωack and the silence time interval Ωsik, the attack model for the k-th attack slot, k{1,,Na}, is proposed as follows:

    ˙ζk(t)=(A+BaFk)ζk(t),ζk(tk)=Δzk
    ak(t)={Fkζk(t),tΩack[0,aTy,k((tk+τk))]T,tΩsik
    a(t)=ki=1ai(t),tΩk

    where the design parameters FkR(nu+ny)×nx and ΔzkRnx are discussed in detail in the sequel. In addition, ay,k(t)=[0, I|Ky|]ak(t) and ay,k((tk+τk))=limt(tk+τk)ay,k(tk+τk). The purpose of the value ak(t) for tΩsik is to guarantee the continuity of the output y(t) at the pausing time instant tk+τk. This design of ak(t) guarantees that the generated attacks can pass through some statistical anomaly detectors (such as some detectors in [42] targeting abrupt value jumps) without being detected. The continuity issue and the corresponding design of ak(t) in (8b) arise due to the intermittent attack implementation mode, whereas typical continuous zero-dynamics attacks such as the ones in [9] and [19] do not have this issue.

    The divergence of a(t) generated by attack model (8) can be avoided through the intermittency performing manner described in (5). Hence, compared with the typical zero-dynamics attack in [19], the intermittent integrity attack a(t) generated by attack model (8) can save the adversary’s instantaneous power energy. It must be noted that the initial condition ζk(tk)=Δzk in (8a), where Δzk is not required to be zero, is more practical than the initial conditions used in [9] and [22]. Specifically, the initial condition Δzk in (8a) can represent the difference between the true value of the system state x and the value known by the attacker. In other words, the attack model (8) does not require the knowledge of the true value of the system states, something that is implied by the nonzero Δzk in (8a), whereas the attack models in [9] and [22] require that. In addition, as shown in the following section, similarly with [9] and [22], the intermittent integrity attacks generated by (8) are stealthy with respect to the standard anomaly detector D if Δzk is sufficiently small. We assume that Δzk satisfies the following assumption.

    Assumption 1: There exist two scalars δ_>0 and ˉδ>0 such that the initial condition Δzk in (8a) is bounded as follows:

    δ_|Δzk|ˉδ,k{1,,Na} (9)

    where δ_ and ˉδ are sufficiently small positive scalars that are not required to be known by the defender.

    Remark 1: The lower bound δ_ implies that the attacker does not need to know the true value of the system state, which is a practical assumption since the true value of the system state is hard to be precisely measured in practice due to the presence of measurement noise. The bound δ_ is also not required to be known by the defender in this paper for developing the attack detection methodology. Furthermore, the (sufficiently small) upper bound ˉδ guarantees that the generated intermittent attacks can pass through the anomaly detector D without any alarms, which is analyzed in detail in the next section. In addition, the next section provides guidance on selecting a suitable ˉδ.

    In [19], it is shown that the stealthiness of zero-dynamics attacks can be violated if Δzk is nonzero. However, [19] does not take into consideration the effects of the controller on the convergence of the system outputs in the attack scenario. In this section, both of the nonzero Δzk(t) and the controller u(t) are considered in analyzing the stealthiness of the generated attacks by (8). To this end, a system splitting is presented. During the attack slot Ωk, Wn in (2) is split into

    Wn1:{˙xn1(t)=Axn1(t)yn1(t)=Cxn1(t)
    Wn2:{˙xn2(t)=Axn2(t)+Bun(t)+Dω(t)yn2(t)=Cxn2(t)+v(t),tΩk

    where the initial conditions are xn1(tk)=0 and xn2(tk)=xn(tk). Such a system splitting guarantees that xn(t)=xn1(t)+xn2(t) and yn(t)=yn1(t)+yn2(t). Using a similar splitting approach, W in (1) is split into W1 and W2 where

    W1:{˙x1(t)=Ax1(t)+Baa(t)y1(t)=Cx1(t)+Daa(t)
    W2:{˙x2(t)=Ax2(t)+Bu(t)+Dω(t)y2(t)=Cx2(t)+v(t),tΩk

    where the initial conditions are x1(tk)=Δzk and x2(tk)=x(tk)+Δzk. Such a splitting also guarantees that x(t)=x1(t)+x2(t) and y(t)=y1(t)+y2(t).

    Throughout this paper, the notation Δ is used to represent the change of any variable due to an attack. For example, Δx is the change of xn due to an attack, i.e., Δx=xxn. Then, the stealthiness of the intermittent integrity attacks generated by (8) is presented.

    Theorem 1 (Stealthiness): Consider the weakly unobservable subspace of W1 in (11a) (denoted by V(W1)), the unobservable subspace of the pair (CA,A) (denoted by H), and the largest controlled invariant subspace of W1 contained in H (denoted by V(H)). By letting V0=V(W1)V(H), if the following conditions are satisfied:

    (A+BaFk)V0V0,(C+DaFk)V0=0
    ΔzkV0,k{1,,Na}

    then the change of output y of W in (1) due to the intermittent integrity attack a(t) generated by (8) is written as

    Δy(t)=ki=1Δy2,i(t),tΩk (13)

    where Δy2,i(t) is generated by the following system:

    ΔW2,i:{Δ˙x2,i(t)=(A+BKC)Δx2,i(t),Δx2,i(ti)=ΔziΔy2,i(t)=CΔzi(t),tΩ0i. (14)

    Proof: The proof is presented in Appendix A.

    Remark 2: Comparing the space V0 with the controlled invariant space for generating the continuous zero-dynamics attack in [19], an additional restriction (the subspace V(H)) is introduced to guarantee that Δy is continuous at the time instant tk+τk. Given the observable pair (C,A), V(H) is nonempty if and only if the matrix A is singular.

    Remark 3: According to the incremental system (14) in Theorem 1, the output change Δy2,k due to the k-th attack converges to zero exponentially with the initial nonzero condition Δzk, since A+BKC is a Hurwitz matrix. Hence, we can conclude that the stealthiness of the generated attacks by model (8) under Assumption 1 with respect to the typical anomaly detector D, can be guaranteed by choosing a sufficiently small initial condition Δzk. In addition, a system with all states measurable (i.e., C is column full rank) does not possess a weakly unobservable subspace V(W1). Hence, no subspace V0 satisfying (12) exists and no stealthy intermittent integrity attack characterized by (8) exists for such a system.

    Theorem 1 shows that Δzk satisfying Assumption 1 causes a nonzero change Δy2,k at the initial time tk, which provides an evidence that the attacks generated by (8) can be detected by some well designed passive analytical redundancy-based detection methodology. Hence, the objective of the rest of this paper is to design a passive attack detection methodology, namely by using only y and u of the system W in (1), for detecting the intermittent integrity attacks generated by (8) under Assumption 1.

    In this section, an equivalent quantity of the system state change at a fixed time prior to the attack occurrence time is introduced, which is referred to as backward-in-time equivalent quantity in this paper. Also, its properties in the context of intermittent integrity attacks, are also rigorously investigated. By using the proposed backward-in-time equivalent quantity, a backward-in-time residual is designed, and the theoretical feasibility of this residual to capture the considered stealthy intermittent integrity attacks is analyzed.

    At first, we suppose that the attacker selects a sufficiently large dwell time τk such that the k-th attack has sufficient time to cause significant damages to the system. Also, the attacker is supposed to set the time length between tk and tk+1 to be sufficiently long such that sufficient energy is recovered for activating the (k+1)-th attack. Then, since the change Δy2,k due to the k-th attack converges to zero with an exponential speed rate (see ΔW2,i in Theorem 1), it is reasonable to consider that Δy2,k(t)0 for ttk+1 in the design of the attack detection methodology. Therefore, based on (13) and (14), we have Δy(t)=Δy2(t), and is described by

    ΔW2:{Δ˙z(t)=(A+BKC)Δz(t),Δz(tk)=ΔzkΔy(t)=CΔz(t),tΩk (15)

    where k{1,,Na}, and Δz(t)=Δx2,k(t) for tΩk. In the rest of this paper, the detection methodology is developed based on system (15).

    Intuitively, the backward-in-time equivalent quantity (mathematically defined later) is a virtual quantity of the system state change, due to the attack, at a time prior to the attack occurrence time, which is recovered from the change of the system state posterior to the attack occurrence time. Based on the backward-in-time equivalent quantity for nonlinear systems given in [36], we define a backward-in-time equivalent quantity of Δz(t) in (15) at a time tb, recovered based on Δz(t) at the time t (ttb).

    Definition 1: The backward-in-time equivalent quantity of the state Δz(t) of the incremental system (15) at a time tb (tbt), denoted by Δz(tb|t), is defined as

    Δz(tb|t)=Φ(tb,t)Δz(t) (16)

    where Φ is the transition matrix associated with A+BKC, i.e.,

    Φ(t1,t2)=e(A+BKC)(t1t2),t1,t2R+. (17)

    The properties of Δz(tb|t) are summarized in the following lemma.

    Lemma 1: Consider the attack generated by (8) satisfying Assumption 1, and the backward-in-time equivalent quantity Δz(tb|t) defined in (16). Then, Δz(tb|t) has the following properties:

    1) The vector Δz(tb|t) is constant for tΩk, i.e.,

    dΔz(tb|t)dt=0,tΩk. (18)

    2) The vector Δz(tb|t) is nonzero for tΩk, i.e.,

    Δz(tb|t)0,tΩk. (19)

    3) (Accumulation property) For the k-th and (k+1)-th attack slots, their recovered backward-in-time equivalent quantities satisfy |Δz(tb|t)|tΩk+1|Δz(tb|t)|tΩk if Δzk+1 and Δzk satisfy the following inequality:

    |Δzk+1|ˉσ(Φ(tb,tk+1))σ_(Φ(tb,tk+1)|Φ(tk+1,tk)Δzk|,k{1,,Na1} (20)

    where the transition matrix Φ is given in (17).

    Proof: The proof is provided in Appendix B.

    It is worth pointing out that under Assumption 1, inequality (20) is easy to satisfy. Due to the exponential convergence of the transition matrix Φ(tk+1,tk) and the sufficiently large time length tk+1tk of the k-th attack slot, Φ(tk+1,tk) is almost zero, so the right hand side of (20) is almost zero as well. Therefore, in the rest of this paper, inequality (20) is considered to hold.

    Lemma 1 implies that the stealthy intermittent integrity attack can be indicated by using a residual based on Δz(tb|t). This is because Δz(tb|t) is monotonically increasing with respect to the attack slot k{1,,Na} given that (20) holds (see result 3)). Next, the backward-in-time residual is proposed by using Δz(tb|t) and is rigorously investigated. Based on Δz(tb|t), the backward-in-time equivalent quantity of Δy(t) in (15) at the time tb is constructed as

    Δy(tb|t)=CΔz(tb|t). (21)

    Then, the backward-in-time residual, denoted by r(tb|t), is defined as follows:

    r(tb|t)=r(tb)+Δy(tb|t) (22)

    where the value r(tb) is given in (3). The threshold with respect to r(tb|t) is chosen as Jth in (4). Then, by using the residual evaluation J(tb|t)=|r(tb|t)|, the occurrence of an anomaly (fault or attack) is ascertained if at some t>T0, J(tb|t) exceeds the threshold Jth, namely

    J(tb|t)>Jth,alarmtriggering. (23)

    Note that the backward-in-time residual r(tb|t) can indicate both faults and attacks, whereas in this paper, we mainly focus on the characteristics of r(tb|t) in terms of detecting the considered stealthy intermittent integrity attacks.

    The detection methodology that integrates the residual r(t) in (3) and the newly proposed backward-in-time residual r(tb|t) in (22), is referred to as ideal backward-in-time detection methodology in this paper. The terminology ideal is used since r(tb|t) is unknown due to the unknown Δy(tb|t). For theoretical analysis purposes at this stage, we consider r(tb|t) to be known. The ideal backward-in-time detection methodology can be found in the left hand side of Fig. 3, which includes the residual generations of r(t) and r(tb|t), their evaluations, and the anomaly detection decision logic. The theoretical feasibility of r(tb|t) to detect stealthy intermittent integrity attacks is presented in the following theorem.

    Figure  3.  Schematic diagram of the ideal and implementable backward-in-time detection methodologies.

    Theorem 2: Consider system (1), the intermittent integrity attack generated by (8) with Δzk satisfying Assumption 1, and a fixed time instant tb. Then, residual r(tb|t) in (22), its evaluation J(tb|t)=|r(tb|t)|, and threshold Jth in (4) satisfy:

    1) In the absence of the intermittent integrity attack

    J(tb|t)Jth,ttb. (24)

    2) In the presence of the intermittent stealthy integrity attack, J(tb|t)>Jth for all tNai=kΩi if condition (20) holds and the k-th attack slot satisfies

    tktb>1λ0lnσ_(C)δ_k0(Jth+|r(tb)|) (25)

    where λ0>0 and k0>0 are scalars satisfying |Φ(t1,t2)| k0eλ0(t1t2) for any t1t2.

    Proof: The proof can be found in Appendix C.

    Remark 4: Theorem 1 and Lemma 1 provide theoretical results for the ideal case that r(tb|t) is known. These theoretical findings prove rigorously that the backward-in-time residual r(tb|t) is able to trigger alarms in the presence of a stealthy intermittent integrity attack generated by (8) under Assumption 1. Result 2) in Theorem 1 is derived from result 3) in Lemma 1. Intuitively, these results imply that during an intermittent integrity attack event with various attack slots, |Δz(tb|t)| increases each time a new attack occurs, and maintains its new value until the next attack occurs (corresponding to (20)). As the number of intermittent attacks accumulates, at some attack occurrence time instant, |r(tb|t)| (see (21) and (22)) exceeds the detection threshold Jth and thus, the intermittent integrity attack is successfully detected (corresponding to (25)).

    In practice, the backward-in-time residual r(tb|t) is unknown due to the unknown Δy(tb|t) (see (21) and (22)). Therefore, one task of the next section is to implement a procedure for estimating optimally the unknown equivalent quantity Δy(tb|t) so that the residual r(tb|t) given by (22) can be implemented in practice and accordingly, a new adaptive threshold is developed based on the implemented estimation procedure.

    In this section, the implementable backward-in-time detection methodology is designed. The right hand side of Fig. 3 shows the structure of the detection methodology. The optimal fixed-point smoother in Fig. 3 is first developed to estimate the unknown Δy(tb|t). The new backward-in-time residual ˆr(tb|t) and the corresponding adaptive threshold ˆJth(t) are then formulated based on the estimation results provided by the smoother. The details are given in the sequel.

    A fixed-point smoother provides a backward-in-time estimation procedure, which produces an estimate for a signal using the past time measurements at the first stage, and then updates it using the new measurements as time progresses. Next, a fixed-point smoother in a finite time horizon [tb,T] is designed for estimating Δy(tb|t). Note that the fixed point tb can be arbitrarily selected by the defender, and so it is available in the detector design.

    Recalling the system splitting given in (11), in the nominal phase (i.e., t<T0), we have x(t)=z(t) and y(t)=y2(t) where z(t) and y2(t) are the state and output of W2 given in (11) respectively. In addition, in the attack phase, Δy(t) equals the output change of W2 (i.e., Δy(t)=Δy2(t), see Theorem 1 and ΔW2 in (15)), and therefore, the smoother is designed based on the system W2 in (11). Let z(tb|t) denote the backward-in-time equivalent quantity of z(t). Then, we have

    z(tb|t)=Δz(tb|t)+z(tb). (26)

    Thus, instead of estimating Δz(tb|t) directly, we first design a fixed-point smoother to estimate z(tb|t) and then, we use relation (26) to reconstruct Δz(tb|t) and relation (21) to reconstruct Δy(tb|t).

    Now, we start by constructing the fixed-point smoother using the state augmentation approach given in [43]. To perform this task, a new state variable ϕ(t)=z(tb|t) for ttb is introduced, where, from result 1) in Lemma 1 and (26), ϕ(t) satisfies

    dϕ(t)dt=˙ϕ(t)=0,tΩk. (27)

    By letting ˆϕ(t) be the estimate of ϕ(t), it follows from (40) and (26) that:

    Δˆy(tb|t)=Cˆϕ(t)y(tb). (28)

    Thus, based on (22), the new backward-in-time residual, denoted by ˆr(tb|t), is proposed as

    ˆr(tb|t)=r(tb)+Δˆy(tb|t). (29)

    Motivated by the optimal residual design methodology in [29], ˆr(tb|t) is to be optimized for achieving robustness with respect to the disturbance ω and the measurement noise v, and sensitivity with respect to the system changes due to attacks. In the sequel, the specific objectives associated with the robustness and sensitivity are presented.

    1) Robustness: The robustness considered in this paper is achieved by minimizing a piece-wise linear quadratic (LQ) cost function. Suppose that a set of smoother switching time instants have been determined, which are given as follows:

    ts,0,ts,1,,ts,Ns+1,NsN+ (30)

    where ts,0=tb and ts,Ns+1=T. Correspondingly, a bank of covariance matrices are introduced and given as follows:

    ˉΘ0,ˉΘ1,,ˉΘNs (31)

    where ˉΘkR2n×2n and ˉΘk0 for all k{0,,Ns}. The objective function associated with robustness is presented in the context of linear fractional transformation (LFT) (see LFT in [44]) in the sequel.

    Consider the system W2 in (11b), and let ˆz be the estimate of z, and ˆy(tb|t)=Cˆϕ(t) be the estimate of y(tb|t) in (21). Then, the smoother design is formulated to find a system K such that the LFT Fl(P,K) can satisfy the robustness requirement where Fl(P,K) is given by

    [˙q(t)y(tb|t)ˆy(tb|t)y(t)]=P[q(t)[ω(t)v(t)]ˆy(tb|t)],ˆy(tb|t)=Ky(t). (32)

    In the above system, q(t)=[zT(t),ϕT(t)]T, P represents the system from [ωT,vT]T and ˆy(tb|t) to y(tb|t)ˆy(tb|t) and y, which can be written in the following matrix form:

    P=[ˉA[ˉD0]0ˉLˉC[00]I[0I]0]

    with ˉA=diag(A,0),ˉC=[C,0],ˉL=[0,C] and ˉD=[DT,0]T. Note that the 2-norm of a system and the 2-norm of its adjoint are equal and also, note that the adjoint of Fl(P,K) is Fl(P,K), in which the notation () represents the adjoint. The adjoint Fl(P,K) is given as follows:

    [dp(τ)dτ˜y(tb|τ)˜ω(τ)]=P[p(τ)˜ω(τ)˜u(τ)],˜u(τ)=K˜ω(τ) (33)

    where τ is a time-to-go variable τ=T+tbt with p(tb)=0, and P is given as follows:

    P=[ˉATˉLTˉCT[ˉDT0]0[00][0I]I0].

    Hence, the piece-wise LQ cost function is readily proposed in the context of the adjoint system (33) as follows:

    minKJ=12Nsk=0p(τs,k)2ˉΘk+12Nsk=0τs,kτs,k+1˜y(tb|τ)2Rdτ (34)

    where τs,k=T+tbts,k for k{0,,Ns+1}, R>0, and the weighting matrix ˉΘk is given in (31) and satisfies some structural requirements. Let Pk and Ωk denote the covariance matrices of z(ts,k)ˆz(ts,k) and ϕ(ts,k)ˆϕ(ts,k) respectively, and Σk denote the mean value of the cross term (z(ts,k)ˆz(ts,k))(ϕ(ts,k)ˆϕ(ts,k))T. Then, Pk>0 and Ωk>0, and the covariance matrix ˉΘk has the following structure:

    ˉΘk=[PkΣkΣTkΩk],k{0,1,,Ns}. (35)

    Note that by choosing ˆz(tb)=ˆϕ(tb)=0, z(tb)ˆz(tb)=ϕ(tb)ˆϕ(tb), P0, Σ0 and Ω0 satisfy

    P0=Σ0=Ω0. (36)

    Furthermore, note also that the estimation accuracy for ϕ is higher than the one for z due to the smoothing process (see [45], [46]). Therefore, the covariance matrix of ϕ(ts,k)ˆϕ(ts,k) is smaller than the one of z(ts,k)ˆz(ts,k), namely, Pk and Ωk must satisfy

    PkΩk,k{1,,Ns}. (37)

    2) Sensitivity: The H performance is introduced to quantitatively scale the sensitivity to attacks. Based on the definition of H in [47], the H performance from zˆz to ˆy(tb|t) is defined as follows:

    H=infzˆzTtb|ˆy(tb|t)|2dtTtb|z(t)ˆz(t)|2dt,z(t)ˆz(t)L2[tb,T]. (38)

    Thus, to guarantee the sensitivity requirement, the inequality

    Hα,z(t)ˆz(t)L2[tb,T] (39)

    must be satisfied, where α>0 is a user-defined performance goal.

    Remark 5: It is worth pointing out that typical optimal LQ fixed-point smoothers in [45] and [46] are obtained by minimizing the following LQ cost function given in the context of the adjoint system (33):

    minKJ1=12p(tb)2ˉΘ0+12Ttb˜y(tb|τ)2Rdτ.

    Such typical LQ smoothers possess inherently the covariance matrix wind-up problem for estimating y(tb|t). As a result, the sensitivity from zˆz to ˆy(tb|t) vanishes as time proceeds, thereby resulting in a conflict to the H performance requirement in (39). More technical details in terms of this issue are given later. In this work, the fixed-point smoother, designed by minimizing the piece-wise LQ cost function (34), can reset the covariance matrix ˉΘk, and is able to guarantee the sensitivity requirement (39) by designing suitable covariance matrices ˉΘk for all k{0,1,,Ns}, i.e., Pk, Σk and Ωk in (35).

    Remark 6: The adjoint system (33) facilitates the presentation of the piece-wise LQ cost function given by (34). Note that ˉΘk, with the structure in (35), is a nonnegative definite matrix and nonsingular, which cannot be used for weighting the initial condition (see [48]). However, in the context of the adjoint system (33), the penalty for the initial condition becomes a penalty for the terminal condition (see the term 12Nsk=0p(τs,k)2ˉΘk in (34)), and in addition, the weighting matrix is not required to be invertible. Hence, the nonsingular matrix ˉΘk in (35) can be used as the weighting matrix in the context of the adjoint system (33).

    By synthesizing the objectives (34) and (39), and the requirements for the covariance matrices given in (35)−(37), a feasible way to solve the optimization problem is given in the following steps:

    1) Minimize J for any ˉΘk0;

    2) Restrict Pk, Σk and Ωk for all k{0,,Ns} such that (36), (37) and (39) are satisfied.

    In the sequel, two lemmas are rigorously derived to realize the aforementioned Steps 1) and 2) respectively. The optimal solution to (34), i.e., Step 1), is first presented.

    Lemma 2: Consider linear system W2 in (11b) and state ϕ in (27). An optimal fixed-point smoother that minimizes (34) in the finite horizon [tb,T] is described by the following dynamics:

    ˙ˆz(t)=Aˆz(t)+Bu(t)+P(t)CTR1(y(t)Cˆz(t))
    ˙ˆϕ(t)=ΣT(t)CTR1(y(t)Cˆz(t))

    where ˆz(tb)=0 and ˆϕ(tb)=0. The matrices P(t)=PT(t)>0, Σ(t) and an auxiliary matrix Ω(t)=ΩT(t)>0 are obtained from the solution of the following differential equations:

    ˙P(t)=AP(t)+P(t)ATP(t)CTR1CP(t)+DDT
    ˙Σ(t)=(AP(t)CTR1C)Σ(t)
    ˙Ω(t)=ΣT(t)CTR1CΣ(t).

    In the above differential equations, at each time ts,k with k{0,,Ns}, the following switch occurs:

    P(ts,k)=Pk,Σ(ts,k)=Σk,Ω(ts,k)=Ωk. (42)

    Moreover, if (37) holds, then the matrices P and Ω satisfy

    P(t)Ω(t),t[tb,T]. (43)

    Proof: Consider the adjoint system (33) and the cost function J in (34). According to LQ control theory (see Section 5.3 in [44]), the cost function J is minimized by the following optimal control law ˜u(τ):

    dˆp(τ)dτ=(ˉATˉCTR1ˉCˉΘ(τ))ˆp(τ)+ˉLT˜ω(τ),ˆp(tb)=0˜u(τ)=R1ˉCˉΘ(τ)ˆp(τ)

    where the covariance matrix ˉΘ(τ) is generated by

    dˉΘ(τ)dτ=ˉAˉΘ(τ)+ˉΘ(τ)ˉATˉΘ(τ)ˉCTR1ˉCˉΘ(τ)+ˉDˉDT

    with the following switches:

    ˉΘ(τs,k)=ˉΘk,k={0,1,,Ns}.

    Then, the optimal smoother is obtained as the adjoint system of ˜u(τ) and is given as follows:

    ˙ˆq(t)=(ˉAˉΘ(t)ˉCTR1ˉC)ˆq(t)+ˉΘ(t)ˉCTR1y(t)
    ˆy(tb|t)=ˉLˆq(t)

    where ˆq=[ˆzT,ˆϕT]T and ˆq(tb)=0, and the covariance matrix ˉΘ(τ) is generated by

    ˙ˉΘ(t)=ˉAˉΘ(t)+ˉΘ(t)ˉATˉΘ(t)ˉCTR1ˉCˉΘ(t)+ˉDˉDT (45)

    with the following switches:

    ˉΘ(ts,k)=ˉΘk,k{0,1,,Ns}. (46)

    Thus, from (44), the smoother (40) can be obtained.

    We now proceed to derive the differential equations given in (41) and switches in (42). By letting

    ˉΘ(t)=[P(t)Σ(t)ΣT(t)Ω(t)].

    It follows from (45) that:

    [˙P˙Σ˙ΣT˙Ω]=[AP(t)CTR1C0ΣT(t)CTR1C0][PΣΣTΩ]+[PΣΣTΩ][ATCTR1CP(t)CTR1CΣ(t)00]+[DP(t)CTR10ΣT(t)CTR1][IR][DTR1CP(t)0R1CΣ(t)].

    By simplifying the above differential Riccati equation, the differential equations (41) can be obtained. In addition, from (35) and (46), the switches in (42) can also be obtained.

    Regarding the result (43), by letting Y(t)=P(t)Ω(t) and AY=AΩCTR1C12YCTR1C, it follows from (41a) and (41c) that:

    ˙Y=AY+YAT+AΩ+ΩATYCTR1CYΩCTR1CYYCTR1CΩYCTR1CΩ+ΩCTR1CΩ+DDT+ΣT(t)CTR1CΣ(t)=AYY+YATY+DDT+ΣT(t)CTR1CΣ(t)+[IΩ][0AATCTR1C][IΩ].

    The above equation indicates that

    ˙Y(t)AYY(t)+YATY(t),t[ts,k,ts,k+1),k{0,,Ns}.

    Under the condition (37), we can obtain that Y(ts,k)0. Therefore, based on Theorem 4.1.2 in [49], Y(t)0 for t[ts,k,ts,k+1) and k{0,,Ns} can be obtained and the result (43) follows.

    It can be observed from (40b) and (41b) that in the case of a typical LQ smoother (obtained by minimizing the cost function J1 and without matrix resetting (42)), the sensitivity from ˆy(tb|t) to zˆz, characterized by the matrix Σ(t), decreases as time proceeds. The reason for this is that the solution Σ(t) of the differential equation (41b) converges to zero as the time progresses, which is the aforementioned covariance matrix wind-up problem. The matrix resetting at the time instant tk characterized in (42) provides an alternative way to maintain the sensitivity. In the following lemma, feasible resetting matrices Pk, Σk and Ωk that satisfy (36), (37) and (39) are presented.

    Lemma 3: Consider the piece-wise fixed-point smoother given in Lemma 2 and the switching time instants given in (30). Consider also the differential equations (41) without the switches (42). The requirements (36) and (37) are guaranteed if

    Pk=P(ts,k),Σk=Σ(ts,k)=Θk,Ωk=Ω(ts,k),k{0,,Ns} (47)

    where P(ts,k), Σ(ts,k) and Ω(ts,k) are the values at ts,k of the solutions to the differential equations (41) without the switches (42) and under the following initial conditions:

    P(tb)=Σ(tb)=Ω(tb)=Θk,k{0,,Ns}. (48)

    Moreover, the H performance index requirement in (39) is guaranteed if Θk in (48) satisfies

    σ_(Θk)ασ_(CR1CT)σ_(C)σ_(e(AP(ts,k+1)CTR1C)(ts,k+1tb)),k{0,1,,Ns}. (49)

    Proof: According to (47), Pk and Ωk are the matrices associated with a typical LQ fixed-point smoother without the matrix resetting. Hence, (37) can be guaranteed directly.

    According to (40b), we can obtain that

    Ttb|ˆy(tb|t)|2dtσ_2(CΣT(t)CTR1C)|z(t)ˆz(t)|2

    which indicates that the H performance requirement in (39) can be guaranteed if

    σ_2(CΣT(t)CTR1C)α2,t[tb,T].

    Based on the inequality σ_(CΣT(t)CTR1C)σ_(CTR1C)× σ_(Σ(t))σ_(C), a sufficient condition to guarantee the above inequality is obtained as

    σ_2(Σ(t))α2σ_2(CTR1C)σ_2(C),t[tb,T]. (50)

    Let X(t)=Σ(t)ΣT(t) where X(tb)=ΘkΘTk. Then, it follows from (41b) without the switches (42) that:

    ˙X(t)=˙Σ(t)ΣT(t)+Σ(t)˙ΣT(t)=(AP(t)CTR1C)X(t)+X(t)(AP(t)CTR1C)T

    where the system ˙x=(AP(t)CTR1C)x is exponentially stable. Moreover, X(t) can be written as

    X(t)=e(AP(t)CTR1C)T(ttb)X(tb)e(AP(t)CTR1C)(ttb) (51)

    which indicates that X(t) is monotonically decreasing with respect to time and X(t)X(tb)=ΘkΘTk for ttb. Thus, the inequality (50) can be guaranteed, if at the end of k-th time interval (i.e., ts,k+1), the following condition is satisfied:

    σ_(X(ts,k+1))α2σ_2(CTR1C)σ_2(C),k{0,1,,Ns}.

    It then follows from (51) and X(tb)=ΘkΘTk that the above inequality can be guaranteed by (49).

    Subsequently, by synthesizing the results in Lemmas 2 and 3, a feasible solution to minimize J in (34) and to satisfy the restrictions (36), (37) and (39) is presented by the following theorem.

    Theorem 3: Consider linear system W2 in (11b) and state ϕ in (27). The fixed-point smoother (40) with P(t), Σ(t) and Ω(t) determined by (41) and the matrix resetting given by (42), minimizes the cost function J in (34) in the finite time horizon [tb,T]. Moreover, by constructing Pk, Σk and Ωk as in (47) and (48), and by choosing Θk to satisfy (49) for k{0, 1,,Ns}, the H performance requirement in (39) and the requirements (36) and (37) are guaranteed simultaneously.

    In this section, the detection residual ˆr(tb|t) in (29) is evaluated and an adaptive threshold is generated. We start by evaluating ˆr(tb|t). Let ez(t)=z(t)ˆz(t), eϕ(t)=ϕ(t)ˆϕ(t) and ey(t)=y(tb|t)ˆy(tb|t) be the estimation errors of x, ϕ and y(tb|t), respectively. Then, from (11b), (27) and (40), the error system is obtained as follows:

    ˙ez(t)=(AP(t)CTR1C)ez(t)+Dω(t)P(t)CTR1v(t)
    ˙eϕ(t)=ΣT(t)CTR1Cez(t)ΣT(t)CTR1v(t)
    ey(t)=Ceϕ(t)

    where the initial conditions are ez(tb)=z(tb) and eϕ(tb)=ϕ(tb). In addition, from ˆy(tb|t)=y(tb|t)ey(t), Δˆy(tb|t)=ˆy(tb|t)y(tb) and Δy(tb|t)=y(tb|t)y(tb), ˆr(tb|t) in (29) can be split into

    ˆr(tb|t)=r(tb)+Δy(tb|t)ey(t). (53)

    As in the case of fault diagnosis literature, the residual evaluation should ensure that in the non-attack case, the value of the evaluation function is close or equal to zero under ideal conditions (e.g., no disturbance, no noise and no modeling uncertainty). Thus, from (53), the evaluation function is proposed as follows:

    ˆJ(tb|t)=|ˆr(tb|t)||r(tb)| (54)

    where the correction term |r(tb)| guarantees that ˆJ(tb|t) is close to zero in the absence of attacks.

    Next, a bound ˆJth(t) (adaptive threshold) of the residual evaluation function ˆJ(tb|t) is derived. To this end, a lemma required to generate the threshold is given.

    Lemma 4: Let Φk(t,τ) be the transition matrix associated with the matrix AP(t)CTR1C for t,τ[ts,k,ts,k+1). Then, there exist known scalars βk>0 and λk>0 such that

    |Φk(t,τ)|βkeλk(tτ),t,τ[ts,k,ts,k+1). (55)

    Proof: Note that since the pair (A,D) is stabilizable and (C,A) is observable, the system ˙x=(AP(t)CTR1C)x is exponentially stable. Thus, x(t)=Φk(t,τ)x(τ) converges exponentially to zero during the time interval t[ts,k,ts,k+1). Hence, such scalars βk and λk that satisfy (55) exist.

    In addition, the disturbance ω(t) and the measurement noise v(t) are supposed to satisfy the following assumption.

    Assumption 2: The disturbance ω(t) and the measurement noise v(t) are bounded by ˉω>0 and ˉv>0, respectively, i.e.,

    |ω(t)|ˉω,|v(t)|ˉv,tR+ (56)

    where ˉω and ˉv are known scalars by the defender.

    Remark 7: Such an assumption is commonly used in anomaly diagnosis literature (see, e.g., [29], [32], [40]) for guaranteeing robustness and avoiding false alarms. The bound ˉω can be obtained a priori by the defender by some experimental tests to the considered CPS, while ˉv can be obtained by exploiting a priori knowledge of the sensor bias deviation based on the technical characteristics of the sensors.

    In the following theorem, the adaptive threshold is presented.

    Theorem 4 (Robustness): Consider closed-loop CPS W in (1) and anomaly detector described in (3) and (4). Also, consider the piece-wise fixed-point smoother (40) described in Theorem 3, the residual (29) and the residual evaluation function (54). Under Assumptions 1 and 2, and in the non-attack case (no intermittent integrity attacks), the residual evaluation function ˆJ(tb|t) in (54) is bounded by the adaptive threshold ˆJth(t) as follows:

    ˆJ(tb|t)ˆJth(t),t[ts,k,ts,k+1) (57)

    where ˆJth(t)=|C|εϕ,k(t) with

    εϕ,k(t)=εϕ,k1(ts,k)+|Σk||CTR1|βktts,keλk(τts,k)(εz,k(τ)+ˉv)dτ (58)
    εz,k(t)=βkeλk(tts,k)|ez(ts,k)|+βk(|D|ˉω+|Pk||CTR1|ˉv)λk(1eλk(ts,kt)). (59)

    In the above equations, βk and λk are specified in Lemma 4.

    Proof: Based on (53) and by using the triangle inequality, the evaluation ˆJ(tb|t) in (54) satisfies

    ˆJ(tb|t)|ey(t)|+|Δy(tb|t)|. (60)

    Note that in the non-attack case, Δz(tb|t)=0 and Δy(tb|t)=0. It then follows from (60) that the threshold ˆJth(t) is chosen as:

    ˆJth(t)=supΔy(tb|t)=0ˆJ(tb|t)=supΔy(tb|t)=0|ey(t)|.

    By solving the differential equation (52a), ez(t) can be explicitly expressed as

    ez(t)=Φk(t,ts,k)ez(ts,k)+tts,kΦk(t,τ)(Dω(τ)P(τ)CTR1v(τ))dτ,t[ts,k,ts,k+1).

    Note that based on Theorem 3.1.1 in [44], P(t) in (41a) is monotonically decreasing with respect to t and thus, P(t)P(ts,k)=Pk for t[ts,k,ts,k+1). Thus, based on Assumption 2 and Lemma 4, we can obtain

    |ez(t)|βkeλk(tts,k)|ez(ts,k)|+tts,kβkeλk(tτ)(|D|ˉω+|Pk|×|CTR1|ˉv)dτ=βkeλk(tts,k)|ez(ts,k)|+βk(|D|ˉω+|Pk|×|CTR1|ˉv)λk(1eλk(ts,kt)).

    Thus, εz,k() in (59) is obtained. In addition, by solving the differential equation (52b), we can obtain

    eϕ(t)=eϕ(ts,k)tts,kΣT(τ)CTR1(ez(τ)+v(τ))dτ,t[ts,k,ts,k+1).

    Note that it follows from (51) that Σ(t) satisfies:

    |Σ(t)|=|X(t)|=|X(ts,k)|×|Φk(t,ts,k)||Σk|βkeλk(tts,k),t[ts,k,ts,k+1).

    Thus, from |ez(t)|εz,k() and |v(t)|ˉv in Assumption 2, and by using the triangle inequality, we can obtain

    |eϕ(t)||eϕ(ts,k)|+|Σk|βktts,keλk(τts,k)|CTR1|(εz,k(τ)+ˉv)dτ.

    Since |eϕ(ts,k)|εϕ,k1(ts,k), then εϕ,k() in (58) is obtained. Hence, from |ey||C||eϕ(t)|, (57) is obtained.

    Remark 8: The threshold ˆJth(t) in (57) is calculated iteratively since εz,k and εϕ,k for the time interval [ts,k,ts,k+1) rely on ez(ts,k) and εϕ,k1(ts,k) respectively from the previous time interval [ts,k1,ts,k). In addition, εz,k in (59) cannot be used directly since |ez(ts,k)| is not available to the defender. To overcome this, it is reasonable to suppose that there exists a scalar δ0>0 such that |ez(ts,k)|δ0. Such δ0 does not affect significantly the final detection result since the term βkeλk(tts,k)|ez(ts,k)|βkeλk(tts,k)δ0 converges to zero exponentially. Therefore, the designer can select a sufficiently large δ0 for implementing the threshold ˆJth(t).

    Based on the residual ˆr(tb|t) in (29), the evaluation function ˆJ(tb|t) in (54), and the threshold ˆJth(t) in (57), the attack occurrence decision principle is given as follows: if there exists a time t>tb such that ˆJ(tb|t) in (54) exceeds the threshold ˆJth(t), i.e., ˆJ(tb|t)>ˆJth(t), then an alarm is triggered to indicate the presence of an attack. The detection time Td of the attack is defined as the first time instant when the inequality ˆJ(tb|t)>ˆJth(t) holds for a given tb, i.e.,

    Td(tb)=inf{t>tb|ˆJ(tb|t)>ˆJth(t)}. (61)

    In addition, as Fig. 3 shows, by combining the residuals r(t) and ˆr(tb|t), the evaluation functions J(t) and ˆJ(tb|t), and the thresholds Jth and ˆJth(t), the occurrence of an anomaly (fault or attack) is decided if there exists a time t>tb such that J(t)>Jth or \hat{J}\left(t_b|t\right)> \hat{J}_{th}(t) .

    Algorithm 1 provides in concise form the steps required for implementing the smoother, generating the backward-in-time residual \hat{r}(t_b|t) and threshold \hat{J}_{th}(t) , and the decision principle for detecting the considered stealthy intermittent integrity attacks.

    Algorithm 1 Backward-in-Time Attack Detection Algorithm

    1: procedure SMOOTHER( t_b , T, t_{s,k} , \alpha , R)   // Theorem 3;

    2:    k\gets0 ;

    3:   repeat        // Lemma 3;

    4:    Choose \Theta_{k} based on (49);

    5:     P(t_b)\gets\Theta_{k} , \Sigma(t_b)\gets\Theta_{k} , \Omega(t_b)\gets\Theta_{k} ;   // (48);

    6:     P_k\gets P(t_{s,k}) , \Sigma_k\gets\Sigma(t_{s,k}) , \Omega_{k}\gets\Omega(t_{s,k}) ; //solve the differential equations in (41); return P_k , \Sigma_k , \Omega_{k} ;

    7:     k\gets k+1 ;

    8:   until k=N_s

    9:    P(t_{s,0})\gets\Theta_{0} , \Sigma(t_{s,0})\gets\Theta_{0} , \Omega(t_{s,0})\gets\Theta_{0} ; // Lemma 2;

    10:   Solve differential equations in (41) for t\in[t_{s,0},t_{s,1}) ;

    11:   return P(t) , \Sigma(t) , \Omega(t) for t\in[t_{s,0},t_{s,1}) ;

    12:    k\gets1 ;

    13:   repeat

    14:     P(t_{s,k})\gets P_{k} , \Sigma(t_{s,k})\gets\Sigma_{k} , \Omega(t_{s,k})\gets\Omega_{k} ; // (42);

    15:    Solve differential equations in (41) for t\in[t_{s,k},t_{s,k+1}) ;

    16:    return P(t) , \Sigma(t) , \Omega(t) for t\in[t_{s,k},t_{s,k+1}) ;

    17:     k\gets k+1 ;

    18:   until k=N_s

    19:   Construct the smoother as follows:     // (40);

        \begin{align*} \dot {\hat {\textit{z}}}(t) & = A\hat {\textit{z}}(t) + Bu(t) + P(t)C^TR^{-1}\left( {y(t) - C\hat {\textit{z}}(t)} \right) \\ \dot {\hat \phi}(t) & = \Sigma^T(t) C^TR^{-1}\left( {y(t)- C\hat {\textit{z}}(t)} \right) \end{align*}

    20: end procedure

    21: //

    22: procedure RESIDUAL( r(t_b),y(t_b),\hat{\phi}(t) )

    23:   Residual \hat{r}\left(t_b|t\right)= r(t_b)+C\hat {\phi} (t)-y(t_b) ;   // (29) and \Delta\hat{y}({t_b}|t)= C\hat{\phi}(t)-y(t_b) ;

    24:   Evaluation \hat{J}\left(t_b|t\right)= |\hat{r}\left(t_b|t\right)|-|r\left(t_b\right)| ;   // (54);

    25: end procedure

    26: //

    27: procedure THRESHOLD( P_k,\Sigma_k,\bar{\omega},\bar{v} )    // Theorem 4;

    28:    k\gets0

    29:   repeat

    30:    Determine \beta_k , \lambda_k satisfying (55); // Lemma 4;

    31:    Determine \delta_0 satisfying |e_{\textit{z}}(t_{s,k})|\leq \delta_0 ;

    32:    Calculate \varepsilon_{z,k}(\cdot) , \varepsilon_{\phi,k}(\cdot) for t\in[t_{s,k},t_{s,k+1}) ; // (59) and (58);

    33:   until k=N_s

    34:   Threshold \hat{J}_{th}(t)=|C|\varepsilon_{\phi,k}(t) for t\in[t_{s,k},t_{s,k+1}) ; // (57);

    35: end procedure

    36: //

    37: procedure DECISION PRINCIPLE( \hat{J}(t_b|t),\hat{J}_{th}(t) )

    38:   if \hat{J}(t_b|t)>\hat{J}_{th}(t) then an alarm is triggered;

    39:   else no attack is detected;

    40:   end if

    41: end procedure

    In this section, the attack detectability of the developed backward-in-time detection methodology characterized by the residual \hat{r}(t_b|t) in (29), the evaluation function \hat{J}(t_b|t) in (54), and the threshold \hat{J}_{th}(t) in (57), is investigated rigorously, characterizing quantitatively the class of detectable intermittent integrity attacks.

    Theorem 5 (Detectability): Consider closed-loop CPS {\cal{W}} in (1) and anomaly detector described in (3) and (4). The attack detection decision scheme, characterized by the piece-wise fixed-point smoother (40) described in Theorem 3, the residual (29), residual evaluation function (54) and detection threshold (57), guarantees that an intermittent integrity attack generated by the model (8) can be detected at a time T_d\geq T_0>t_b , i.e., \hat{J}({t_b|T_d})>\hat{J}_{th}(T_d) , if for the given fixed time instant t_b , there exists an attack slot k such that

    t_{k}-t_b>+\dfrac{1}{\lambda_{0}}\ln\dfrac{\underline{\sigma}(C)\underline{\delta}}{k_{0}\left(J_{th}+2\hat{J}_{th}(T_d)+|r(t_b)|\right)} (62)

    where \underline{\delta} is given in Assumption 1, k_0 and \lambda_0 are specified in Theorem 2, J_{th} is given in (4) and \hat{J}_{th} is given in (57).

    Proof: For \hat{r}\left(t_b|t\right) = r\left(t_b|t\right)-e_{y}(t) given in (53), by using the reverse triangle inequality, we can obtain

    \begin{align} |\hat{r}(t_b|T_d)|\geq|r(t_b|T_d)|-|e_y(T_d)|. \end{align} (63)

    From (54) and |r(t_b)|< J_{th} , to detect an attack at the time instant T_d , i.e., \hat{J}({t_b|T_d})>\hat{J}_{th}(T_d) , the following inequality must hold:

    |\hat{r}\left(t_b|T_d\right)|>J_{th}+\hat{J}_{th}(T_d). (64)

    Then, from (63) and the fact that |e_{y}(t)|\leq \hat{J}_{th}(t) for t\leq T_d , a sufficient condition to guarantee (64) can be obtained as follows:

    \begin{align} |r(t_b|T_d)|>J_{th}+2\hat{J}_{th}(T_d). \end{align} (65)

    By using the same reasoning logic with the proof of Theorem 2, we can obtain that for the fixed time instant t_b , if there exists an attack slot k satisfying (62), then the inequality (65) can be guaranteed for any T_d\in\bigcup_{i = k}^{N_a}\Omega_i . Hence, the result is proved.

    Theorem 5 is a theoretical result that cannot be checked a priori. It is important to note that according to Theorem 5, the attack detection by the developed implementable backward-in-time detection methodology, characterized by the residual \hat{r}(t_b|t) in (29), the evaluation function \hat{J}(t_b|t) in (54), and the threshold \hat{J}_{th}(t) in (57), is guaranteed by selecting a small t_b (i.e., close to zero) so that condition (62) can hold. In other words, as time progresses and at each attack activation time t_k , the left side of (62) increases and at some time, will exceed the right hand side of (62), leading to the detection of the intermittent integrity attack.

    In this section, a numerical simulation example based on a linear time-invariant system in the form of system (1) is presented. The system matrices are given as follows:

    \begin{split} &A = \left[ {\begin{matrix} -3.25&1&0\\ 1&-3&0\\ 0&0&0 \end{matrix}} \right],\; B = \left[ {\begin{matrix} 0\\ 0\\ 1 \end{matrix}} \right]\;\\ &C = \left[ {\begin{matrix} 0&1&1 \end{matrix}} \right],\; D = \left[ {\begin{matrix} 1.0000&0\\ 0&0\\ 1.6667&1.6667 \end{matrix}} \right] \end{split}

    where the pair ( C,A ) is observable and (A,D) is stabilizable. The output control gain is given by K = -4.7333 and the reference signal y_{ {{\rm{ref}}}} = 0 . In addition, for the simulation purpose, the disturbance \omega(t) is given by

    \omega(t) = [0.2\sin(5t),0.1\sin(3t)]^T

    and the measurement noise v(t) is chosen at each time instant as a uniformly distributed random number ranging from -0.3 to 0.3 . The anomaly detector {\cal{D}} in (3) is designed based on [29]. The residual is designed to satisfy the optimal {\cal{H}}_{\infty} performance, i.e., \int_{0}^{15}r^T(t)r(t)dt\;\leq \;2\int_{0}^{15}[\omega^T(t), v^T(t)]^T[\omega^T(t), v^T(t)]dt and the threshold is chosen as J_{th} = 5 based on the fault and attack free operation.

    In this part, the intermittent integrity attack used in this simulation is given. The attacker is supposed to know the system matrices A, B and C, and is able to compromise all the sensors and actuators, i.e., \Gamma_u = 1 and \Gamma_y = 1 . The attack activating time instants t_k are given first as follows:

    t_1 = 1\, {{\rm{s}}},\;t_2 = 3\, {{\rm{s}}},\;t_3 = 5\, {{\rm{s}}},\;t_4 = 7 \,{{\rm{s}}},\;t_5 = 9 \,{{\rm{s}}},\;t_6 = 11 \,{{\rm{s}}},\;t_7 = 15 \,{{\rm{s}}}

    and the same dwell time is used for all attacks, i.e., \tau_k = 1\;{{\rm{s}}} for all k\in\{1,\ldots,6\} . In the sequel, the design parameters of the attack model (8) are calculated. Based on Theorem 1, and by using the geometric approach toolbox in [50], we can obtain that the subspace {\cal{V}}_0 satisfying Theorem 1 is {\cal{V}}_0 = [0,0,-1]^T and further, a feasible F_k satisfying (12a) is calculated as

    F_k = \left[ \begin{array}{ccc} 0 & 0 & 1.200\\ 0 & 0 & -1.000 \end{array} \right],\; \forall\; k\in\{1,\ldots,6\}.

    According to the obtained {\cal{V}}_0 , \Delta {\textit{z}}_k satisfying (12b) is chosen as

    \Delta {\textit{z}}_1 = [0,0,-0.2091],\; \Delta {\textit{z}}_2 = [0,0,-0.2210]
    \Delta {\textit{z}}_3 = [0,0,-0.1191],\; \Delta {\textit{z}}_4 = [0,0,-0.1531]
    \Delta {\textit{z}}_5 = [0,0,-0.1531],\; \Delta {\textit{z}}_6 = [0,0,-0.1095].

    Thus, the design parameters of the attack model (8) have been selected and the attack signals for the attack activating time interval \Omega^{\mathrm{ac}}_k can be generated. Next, the attack signals during the attack silence time interval \Omega^{\mathrm{si}}_k , i.e., a_{y,k}((t_k+\tau_k)^{-}) , are given based on (8b) as follows:

    \begin{align*} {a_1}\left( t \right)& = {\left[ {0,-1.2919} \right]^T},\; k = 1,\; \forall t \in \left[ {2\,{\rm{s}},15\,{\rm{s}}} \right) \\ {a_2}\left( t \right)& = {\left[ {0,-1.2392} \right]^T},\; k = 2,\; \forall t \in \left[ {4\,{\rm{s}},15\,{\rm{s}}} \right) \\ {a_3}\left( t \right)& = {\left[ {0,-1.2060} \right]^T},\; k = 3,\; \forall t \in \left[ {6\,{\rm{s}},15\,{\rm{s}}} \right) \\ {a_4}\left( t \right)& = {\left[ {0,-1.0931} \right]^T},\; k = 4,\; \forall t \in \left[ {8\,{\rm{s}},15\,{\rm{s}}} \right) \\ {a_5}\left( t \right)& = {\left[ {0,-1.0931} \right]^T},\; k = 5,\; \forall t \in \left[ {10\,{\rm{s}},15\,{\rm{s}}} \right) \\ {a_6}\left( t \right)& = {\left[ {0,-1.2432} \right]^T},\; k = 6,\; \forall t \in \left[ {12\,{\rm{s}},15\,{\rm{s}}} \right). \end{align*}

    Thus, the intermittent integrity attack signal for the attack activating time interval \Omega^{\mathrm{ac}}_k and the attack silence time interval \Omega^{\mathrm{si}}_k is generated.

    The attack signal and its effects on the system are shown in Figs. 4-6, respectively. As it is shown in Fig. 4, the attack signal a_y(t) is continuous at the attack pausing time instants \{2 \,{{\rm{s}}},\;4\, {{\rm{s}}},\;6\, {{\rm{s}}},\;8 \,{{\rm{s}}},\;10\, {{\rm{s}}},\;12\, {{\rm{s}}}\}. Moreover, the resulting system output y and its change in Fig. 6 have no jump (abrupt change) at each of these attack pausing time instants. By comparing the system states x in the attack case with the state x^n in the nominal case in Fig. 5, we can observe that the system state x_3 is significantly affected by the injected intermittent integrity attacks. However, in contrast to y^n in the nominal case, the system output y in Fig. 6 has a very small increment \Delta y at each of the attack resuming time instants \{1\, {{\rm{s}}},\;3\, {{\rm{s}}},\;5 \,{{\rm{s}}},\;7\, {{\rm{s}}},\;9 \,{{\rm{s}}}, 11\, {{\rm{s}}}\}, and such an increment converges to zero exponentially. Fig. 6 also shows that the increments caused by the attack are relatively small and hidden by the disturbances and noise, which is a result of the selected relatively “large” process disturbances and measurement noise used in the simulation. This particularly created simulation scenario is used for verifying that the designed backward-in-time detector is robust to the process disturbances and measurement noise, and also sensitive to the stealthy intermittent integrity attacks.

    Figure  4.  Time responses of the attack signals a_u(t) and a_y(t) .
    Figure  6.  Time responses of the sensor measurements y(t) and y^n(t) in the attack case and nominal case respectively, and the output change \Delta{y}(t) .
    Figure  5.  Time responses of the state vector x(t) in the attack case and x^n(t) in the nominal case.

    Fig. 7 illustrates the anomaly detection results using the equipped anomaly detector {\cal{D}} , in which the residual r(t) remains far below the threshold J_{th} during the attack event. Hence, the attack is not detected by {\cal{D}} . This indicates that the change \Delta y due to the intermittent attack is sufficiently small to maintain the stealthiness of the attack with respect to the detector {\cal{D}} .

    Figure  7.  Time responses of the forward-in-time residual r(t) , the evaluation function J(t) = |r(t)| and the threshold J_{th} of the anomaly detector {\cal{D}} .

    Following Algorithm 1, the parameters for the SMOOTHER procedure are given as follows: t_b = 2,\; T = 15,\; t_{s,0} = t_b = 2,\; t_{s,1} = 9,\; t_{s,2} = 16, \alpha = 0.009 and R = 0.1 . Moreover, \Theta_k satisfying (49) is given by \Theta_{0} = 8.5I_3,\Theta_{1} = 43.5I_3 . By solving the differential equations in (41), P_1 and \Sigma_1 are obtained as follows:

    P_1 = \left[\begin{array}{ccc} 0.0162& 0.0025& 0.0484\\ 0.0027& 0.0008& 0.0153\\ 0.0544& 0.0101& 0.2307 \end{array}\right]
    \Sigma_1 = \left[\begin{array}{ccc} 0.0010& 0& 0\\ 0& 0.0182& 0\\ 0& 0& 29.232\\ \end{array}\right].

    Furthermore, P(t) and \Sigma(t) for [2 \;{{\rm{s}}},\;9 \;{{\rm{s}}}) and [9 \;{{\rm{s}}},\;16\; {{\rm{s}}}) can be obtained respectively by solving the differential equations in (41), and thus, the fixed-point smoother can be implemented. Then, given y(t_b) = 0.1985 , r(t_b) = -0.1332 and the estimate \hat{\phi}(t) , the RESIDUAL procedure can be completed by following Algorithm 1, and the residual \hat{r}(t_b|t) from (29) and its evaluation function \hat{J}(t_b|t) from (54) can be obtained. We proceed with the THRESHOLD procedure based on Algorithm 1. The scalars \bar{\omega} and \bar{v} in Assumption 2 are given by \bar{\omega} = 0.3 and \bar{v} = 0.2 . Moreover, \beta_0 = 0.25 , \lambda_{0} = 0.64 , \beta_1 = 0.1 and \lambda_1 = 0.63 , and \delta_0 = 100 . The initial value of \varepsilon_{\phi,k} is chosen as \varepsilon_{\phi,0} = 5 . By following the THRESHOLD procedure in Algorithm 1, the threshold \hat{J}_{th}(t) in (57) can be calculated.

    The residual \hat{r}(2|t) , the evaluation function \hat{J}(2|t) and the threshold \hat{J}_{th}(t) are shown in Fig. 8. It is shown that at each of the attack resuming time instants \{3\, {{\rm{s}}},\;5\, {{\rm{s}}},\;7 \,{{\rm{s}}},\;9\, {{\rm{s}}},\;11\, {{\rm{s}}}\}, the residual \hat{r}(2|t) and its corresponding evaluation \hat{J}(2|t) have a jump in magnitude. Thus, the accumulation property described in Lemma 1 is satisfied. Furthermore, note also that in Fig. 8, the jump that occurs at t = 9 \;{{\rm{s}}} is much larger than the one at t = 7\; {{\rm{s}}} due to covariance matrix resetting, even though the increment \Delta y at t = 9 \;{{\rm{s}}} is similar with the one at t = 7 \;{{\rm{s}}} (see \Delta {\textit{z}}_4 = \Delta {\textit{z}}_5 ). Moreover, based on the DECISION PRINCIPLE specified in Algorithm 1, we can conclude from Fig. 8 that the injected intermittent stealthy integrity attack is successfully detected at the time T_d\approx9.5 \;{{\rm{s}}} when the evaluation function \hat{J}(2|t) exceeds the threshold \hat{J}_{th}(t) .

    Figure  8.  Time responses of the backward-in-time residual \hat{r}(2|t) , the evaluation function \hat{J}(2|t) and the threshold \hat{J}_{th}(t) .

    In this paper, stealthy intermittent integrity attacks being stealthy with respect to typical anomaly detectors have been formulated. A backward-in-time detection residual that can accumulate at each attack activation time and is able to indicate the stealthy intermittent integrity attacks has been introduced. A fixed-point smoother has been designed as the backward-in-time estimator for estimating the unknown backward-in-time detection residual. A covariance matrix resetting technique has been applied in the design of the smoother to guarantee the required sensitivity to the attacks. The corresponding adaptive threshold generation scheme for detecting the stealthy intermittent integrity attacks has been designed, and the attack detectability has also been investigated rigorously. Some future research works are given as follows:

    1) One of our studies work focuses on scheduling the attack pausing and resuming time instants such that the attack can cause significant damage to the system, and at the same time, achieve the power energy saving aim. Game theory may provide a way to solve the trade-off between damaging effects and energy saving [38].

    2) Another future research direction involves the modification of typical fixed-point smoothers to improve the sensitivity to stealthy integrity attacks. The forgetting factor for the covariance matrix used in [51] is a potential way to achieve this improvement.

    3) Resilience control against intermittent integrity attacks is also one potential research direction. Some control issues, such as quantized sensor measurements well handled by the model reference control methodologies in [52], [53] and high order nonlinearities in [54], [55] will be investigated.

    4) Stealthy intermittent integrity attack generation and detection issues for nonlinear systems and large-scale systems such as in [56] will be considered in our future research. Note that a way for generating stealthy intermittent attacks for nonlinear systems and large-scale systems is to use the geometric approach proposed in [14].

    Proof : Let \Delta x_1 , \Delta x_2 , \Delta y_1 , \Delta y_2 and \Delta u represent the changes of x_1 , x_2 , y_1 , y_2 and u respectively due to the attack a(t) , i.e., \Delta x= x-x^n , \Delta x_2= x_2-x_2^n , \Delta y_1= y_1-y_1^n , \Delta y_2= y_2-y_2^n and \Delta u= u-u^n . Then, from (10) and (11), the incremental systems between {\cal{W}}_{j} and {\cal{W}}^n_{j} for j = 1,2 , denoted as \Delta{\cal{W}}_{j} , can be obtained as follows:

    \begin{aligned} \Delta{\cal{W}}_{1}:&\; \left\{ \begin{aligned}\Delta\dot{x}_{1}(t)& = A\Delta x_{1}(t)+B_aa(t)\\ \Delta y_{1}(t)& = C\Delta x_{1}(t)+D_aa(t)\end{aligned}\right. \end{aligned} (66)
    \begin{align} \Delta{\cal{W}}_{2}:&\; \left\{ \begin{aligned}\Delta\dot{x}_2(t)& = A\Delta x_2(t)+B\Delta u(t)\\ \Delta y_{2}(t)& = C\Delta x_2(t)\end{aligned}\right. \end{align} (67)

    where \Delta x_{1}(t_k) = -\Delta {\textit{z}}_k and \Delta x_2(t_k) = \Delta {\textit{z}}_k . Thus, the state change \Delta x and the output change \Delta y can be written as

    \begin{align} \Delta x(t)& = \Delta x_{1}(t)+\Delta {\textit{z}}(t) \;\;\;\end{align} (68)
    \begin{align} \Delta y(t)& = \Delta y_{1}(t)+\Delta y_{2}(t). \end{align} (69)

    Consider {\cal{W}}_j for j = 1,2 given in (11). Denote {\cal{W}}^n_{j,k} as the “nominal” system of {\cal{W}}_j in the absence of the attack signal a_k(t) but in the presence of a_1(t),\ldots,a_{k-1}(t) , where k\geq1 . By letting x^n_{1,k} and y^n_{1,k} represent the state and the output of {\cal{W}}^n_{1,k} , and letting x^n_{2,k} , y^n_{2,k} and u^n_{k} represent the state, the output and the control of {\cal{W}}^n_{2,k} , respectively, it then follows from (11) that {\cal{W}}^n_{j,k} can be written as:

    \begin{align} {\cal{W}}^n_{1,k}:&\; \left\{ \begin{aligned}\dot{x}^n_{1,k}(t)& = Ax^n_{1,k}(t)+\sum\limits_{i = 1}^{k-1}B_aa_k(t)\\ y^n_{1,k}(t)& = Cx^n_{1,k}(t)+\sum\limits_{i = 1}^{k-1}D_aa_k(t)\end{aligned} \right. \end{align} (70)
    \begin{align} {\cal{W}}^n_{2,k}:&\; \left\{ \begin{aligned} \dot{x}_{2,k}^n(t)& = Ax_{2,k}^n(t)+Bu_{k}^n(t)+D\omega(t)\\ y^n_{2,k}(t)& = Cx_{2,k}^n(t)+v(t) \end{aligned} \right. \end{align} (71)

    where the initial conditions can be chosen as x_{1}(t_k)-x^n_{1,k}(t_k) = -\Delta {\textit{z}}_k and x_2(t_k)-x^n_{2,k}(t_k) = \Delta {\textit{z}}_k such that x_{1}(t_k)+x_2(t_k) = x^n_{1,k}(t_k)+ x^n_{2,k}(t_k).

    Let \Delta x_{1,k} and \Delta y_{1,k} represent the changes of x_{1,k}^n , and y_{1,k}^n , and \Delta x_{2,k} , \Delta y_{2,k} and \Delta u_{k} represent the changes of x_{2,k}^n , y_{2,k}^n and u_{k}^n respectively due to the attack a_k(t) , i.e., \Delta x_{1,k}= x_{1}-x_{1,k}^n , \Delta y_{1,k}= y_{1}-y_{1,k}^n , \Delta x_{2,k}= x_2-x_{2,k}^n , \Delta y_{2,k}= y_{2}-y_{2,k}^n and \Delta u_{k}= u- u_{k}^n. Then, based on (70), (71) and (11), the incremental systems, denoted as \Delta{\cal{W}}_{j,k} , are obtained as

    \begin{align} \Delta{\cal{W}}_{1,k}:&\; \left\{ \begin{aligned}\Delta\dot{x}_{1,k}(t)& = A\Delta x_{1,k}(t)+B_aa_k(t)\\ \Delta y_{1,k}(t)& = C\Delta x_{1,k}(t)+D_aa_k(t)\end{aligned}\right. \end{align} (72)
    \begin{align} \Delta{\cal{W}}_{2,k}:&\; \left\{ \begin{aligned}\Delta\dot{x}_{2,k}(t)& = A\Delta x_{2,k}(t)+B\Delta u_{k}(t)\\ \Delta y_{2,k}(t)& = C\Delta x_{2,k}(t)\end{aligned}\right. \end{align} (73)

    where \Delta x_{1,k}(t_k) = -\Delta {\textit{z}}_k and \Delta x_{2,k}(t_k) = \Delta {\textit{z}}_k .

    Therefore, based on the superposition principle for linear systems ( a(t) is the sum of a_k(t) in (8c)), \Delta x_1 , \Delta {\textit{z}} and \Delta y can be written as

    \begin{align} \Delta x_1(t)& = \sum\limits_{i = 1}^{k} \Delta x_{1,i}(t),\; \Delta x_2(t) = \sum\limits_{i = 1}^{k} \Delta x_{2,i}(t) \end{align} (74)
    \begin{align} \Delta y(t)& = \sum\limits_{i = 1}^{k} \left(\Delta y_{1,i}(t)+\Delta y_{2,i}(t)\right),\; \forall\; t\in\Omega_{k}. \end{align} (75)

    In the sequel, the responses of \Delta{\cal{W}}_{j,k} in the attack activating time interval and the attack silence time interval are analyzed.

    1) Activating Time Interval \Omega^{\mathrm{ac}}_k : In this time interval, \Delta {\cal{W}}_{1,k} and a_k(t) described by (8a) and (8b) respectively can be equivalently written in the coordinates (\Delta\bar{x}_{1,k},\zeta_k) with \Delta\bar{x}_{1,k}= \Delta x_{1,k}-\zeta_k as follows:

    \begin{align*} &{\Delta\dot {\bar{x}}}_{1,k}(t) = A\Delta\bar{x}_{1,k}(t)\\ & {\dot \zeta_k}(t)= ({A + B_aF_k})\zeta_k(t)\\ &\Delta y_{1,k}(t) = C\Delta\bar{x}_{1,k}(t)+(C+D_aF_k)\zeta_k(t) \end{align*}

    where \Delta\bar{x}_{1,k}(t_k) = \Delta x_{1,k}(t_k)-\zeta_k(t_k) = 0 . Since \Delta\bar{x}_{1,k}(t_k) = 0 , \Delta\bar{x}_{1,k}(t) = 0 for t\in\Omega^{\mathrm{ac}}_k , and hence, \Delta x_{1,k}(t) = \zeta_k(t) for t\in\Omega^{\mathrm{ac}}_k . For \Delta {\textit{z}}_{k}\in{\cal{V}}_0 and F_k satisfying (12a), we have

    \zeta_k(t)\in{\cal{V}}_0,\; (C+D_aF_k)\zeta_k(t) = 0,\; \forall\; t\in\Omega^{\mathrm{ac}}_k.

    Thus, we obtain

    \Delta y_{1,k}(t) = 0,\; \forall\; t\in\Omega^{\mathrm{ac}}_k. (76)

    2) Silence Time Interval \Omega^{\mathrm{si}}_k : In this time interval, the initial condition of \Delta {\cal{W}}_{1,k} is \Delta x_{1,k}(t_k+\tau_k) = \Delta x_{1,k}((t_k+\tau_k)^{-})\in{\cal{V}}_0 . Based on a_{k}(t) in (8b) during t\in\Omega^{\mathrm{si}}_k , \Delta x_{1,k} satisfies

    \Delta\dot{x}_{1,k}(t) = A\Delta x_{1,k}(t),\; \Delta x_{1,k}(t_k+\tau_k)\in{\cal{V}}_0.

    Also, \Delta y_{1,k} at the time instant t_k+\tau_k satisfies

    \begin{split} \Delta y_{1,k}(t_k+\tau_k)& = C\Delta x_{1,k}(t)+a_{y,k}((t_k+\tau_k)^-)\\ & = \Delta y_{1,k}((t_k+\tau_k)^-) = 0. \end{split} (77)

    Thus, it follows from \Delta x_{1,k}(t_k+\tau_k)\in{\cal{V}}_0 and {\cal{V}}_0\subset{\cal{H}} with {\cal{H}} being the unobservable subspace of the pair (CA,A) that \Delta x_{1,k}(t_k+\tau_k)\in{\cal{H}} , which indicates:

    \begin{align} \Delta\dot{y}_{1,k}(t) = CA\Delta x_{1,k}(t) = 0,\; \forall\; t\in\Omega^{\mathrm{si}}_k. \end{align} (78)

    Thus, by combining (77) and (78), we have

    \begin{align} \Delta y_{1,k}(t) = 0,\; \forall\; t\in\Omega^{\mathrm{si}}_k. \end{align} (79)

    Hence, from the result (76) in the activating time interval and the result (79) in the silence time interval, we can conclude

    \begin{equation*} \Delta y_{1,k}(t) = 0,\; \forall\; t\in\Omega^0_k. \end{equation*}

    Therefore, it follows from (75) that the change \Delta y can be written as in (13) with \Delta y_{2,k}(t) being generated by \Delta{\cal{W}}_{2,k} in (79). In addition, since \Delta u_k = K\Delta y_k = K\Delta y_{2,k} = KC\Delta {\textit{z}}_k , (79) can be written as (14).

    Proof:

    1) By using the transition matrix Φ in (17), the solution of the system \Delta {\cal{W}} in (15) with the initial condition \Delta {\textit{z}}(t_k) = \Delta {\textit{z}}_{k} can be written as

    \Delta {\textit{z}}(t) = \Phi(t,t_k)\Delta {\textit{z}}_{k},\; \forall\; t\in\Omega_k.

    Thus, from Definition 1, {\Delta {\textit{z}}}(t_b|t) can be written as

    \begin{align} {\Delta {\textit{z}}}(t_b|t) = \Phi({t_b,t})\Phi(t,t_k)\Delta {\textit{z}}_{k} = \Phi(t_b,t_k)\Delta {\textit{z}}_{k},\; \forall\; t\in\Omega_k. \end{align} (80)

    Since both \Phi(t_b,t_k) and \Delta {\textit{z}}_{k} are independent of time t, {\Delta {\textit{z}}}(t_b|t) is also independent of time t, then {\Delta {\textit{z}}}(t_b|t) is a constant vector with respect to time. Hence, (18) follows.

    2) Since \Delta {\textit{z}}_{k}\neq0 and \Phi(t_b,t_k)\neq0 , then the result (19) follows directly from (80).

    3) From (80), for the consecutive attack slots \Omega_k and \Omega_{k+1} , we have

    \begin{align*} \Delta {\textit{z}}(t_b|t)& = \Phi(t_b,t_k)\Delta {\textit{z}}_k,\; \forall\; t\in\Omega_{k}\\ \Delta {\textit{z}}(t_b|t)& = \Phi(t_b,t_{k+1})\Delta {\textit{z}}_{k+1},\; \forall\; t\in\Omega_{k+1}. \end{align*}

    By using \Phi(t_b,t_{k}) = \Phi(t_b,t_{k+1})\Phi(t_{k+1},t_{k}) , \Delta {\textit{z}}(t_b|t_{k}) can be equivalently written as

    \begin{align*} \Delta {\textit{z}}(t_b|t) = \Phi(t_b,t_{k+1})\Phi(t_{k+1},t_{k})\Delta {\textit{z}}_{k},\; \forall\; t\in\Omega_{k}. \end{align*}

    Then, we can derive

    \begin{align*} |\Delta {\textit{z}}(t_b|t)|^2&\leq\bar{\sigma}^2(\Phi(t_b,t_{k+1})|\Phi(t_{k+1},t_{k})\Delta {\textit{z}}_{k}|^2,\; \forall\; t\in\Omega_{k}\\ |\Delta {\textit{z}}(t_b|t)|^2&\geq\underline{\sigma}^2(\Phi(t_b,t_{k+1}))|\Delta {\textit{z}}_{k+1}|^2,\; \forall\; t\in\Omega_{k+1}. \end{align*}

    Hence, |{\Delta {\textit{z}}}(t_b|t)|_{t\in\Omega_{k+1}}\geq|{\Delta {\textit{z}}}(t_b|t)|_{t\in\Omega_{k}} if (20) is satisfied.

    Proof:

    1) Based on Lemma 1, in the absence of the attack, {\Delta {\textit{z}}}(t_b|t) = 0 and from (21), \Delta y({t_b}|t) = 0 , and thus, r\left(t_b|t\right) = r(t_b) . Since |r(t_b)|\leq J_{th} , result 1) follows.

    2) By using the reverse triangle inequality and based on (21) and (22), a sufficient condition to guarantee |r(t_b|t)|> J_{th} for all t\in\bigcup_{i = k}^{N_a}\Omega_i can be obtained as

    |C\Delta {\textit{z}}({t_b}|t)|> \dfrac{J_{th}+|r(t_b)|}{\underline{\sigma}(C)},\; \forall\; t\in\bigcup\limits_{i = k}^{N_a}\Omega_i. (81)

    Note that based on result 1) in Lemma 1, \Delta {\textit{z}}({t_b}|t) is a constant vector during an attack slot \Omega_k . Thus, |\Delta {\textit{z}}({t_b}|t)| = |\Delta {\textit{z}}({t_b}|t_k)| for t\in\Omega_k . Note also that under Assumption 1 and based on result 3) in Lemma 1, |\Delta {\textit{z}}({t_b}|t_i)|\geq|\Delta {\textit{z}}({t_b}|t_k)| for any i\geq k since (20) is considered to hold. Thus, a sufficient condition to guarantee (81) is obtained as follows:

    |\Delta {\textit{z}}({t_b}|t_k)|> \dfrac{J_{th}+|r(t_b)|}{\underline{\sigma}(C)}. (82)

    Based on Definition 1, we can write \Delta {\textit{z}}({t_b}|t_k) = \Phi^{-1}(t_k,t_b)\Delta {\textit{z}}_k . Thus, a sufficient condition to guarantee (82) can be obtained as

    |\Phi(t_{k},t_b)|< \dfrac{\underline{\sigma}(C)|\Delta {\textit{z}}_k|}{J_{th}+|r(t_b)|}. (83)

    Note that according to [40], for the Hurwitz matrix A+BKC , there exist k_0>0 and \lambda_0>0 such that

    \begin{equation*} |\Phi(t_{k},t_b)|\leq k_{0} e^{-\lambda_{0}(t_{k}-t_b)}. \end{equation*}

    Thus, the time t_k satisfying (25) can guarantee the sufficient condition (83). Hence, result 2) follows.

  • 1 Covariance matrix and mean value are concepts used in Kalman filtering. Since LQ optimal filters have similar form with the Kalman filter, we also use the terminologies “covariance matrix” and “mean value” for the LQ optimal filters.2 Regarding adjoint system of a linear system, the definition can be found in [44].
    2 Regarding adjoint system of a linear system, the definition can be found in [44].
  • [1]
    A. A. Cardenas, S. Amin, and S. Sastry, “Secure control: Towards survivable cyber-physical systems,” in Proc. 28th Int. Conf. Distributed Computing Systems Workshops, Beijing, China, 2008, pp. 495–500.
    [2]
    S. M. Dibaji, M. Pirani, D. B. Flamholz, A. M. Annaswamy, K. H. Johansson, and A. Chakrabortty, “A systems and control perspective of CPS security,” Annu. Rev. Control, vol. 47, pp. 394–411, Jan. 2019. doi: 10.1016/j.arcontrol.2019.04.011
    [3]
    V. L. Do, L. Fillatre, I. Nikiforov, and P. Willett, “Security of SCADA systems against cyber-physical attacks,” IEEE Aerosp. Electron. Syst. Mag., vol. 32, no. 5, pp. 28–45, May 2017. doi: 10.1109/MAES.2017.160047
    [4]
    A. Hobbs, “The colonial pipeline hack: Exposing vulnerabilities in U.S. cybersecurity,” 2021. [Online]. Available: https://sk.sagepub.com/cases/colonial-pipeline-hack-exposing-vulnerabilities-us-cybersecurity.
    [5]
    W. L. Duo, M. C. Zhou, and A. Abusorrah, “A survey of cyber attacks on cyber physical systems: Recent advances and challenges,” IEEE/CAA J. Autom. Sinica, vol. 9, no. 5, pp. 784–800, May 2022. doi: 10.1109/JAS.2022.105548
    [6]
    Y. L. Mo and B. Sinopoli, “Secure control against replay attacks,” in Proc. 47th Annu. Allerton Conf. Communication, Control, and Computing, Monticello, USA, 2009, pp. 911–918.
    [7]
    R. S. Smith, “Covert misappropriation of networked control systems: Presenting a feedback structure,” IEEE Control Syst. Mag., vol. 35, no. 1, pp. 82–92, Feb. 2015. doi: 10.1109/MCS.2014.2364723
    [8]
    A. Barboni, H. Rezaee, F. Boem, and T. Parisini, “Detection of covert cyber-attacks in interconnected systems: A distributed model-based approach,” IEEE Trans. Autom. Control, vol. 65, no. 9, pp. 3728–3741, Sept. 2020. doi: 10.1109/TAC.2020.2998765
    [9]
    A. Teixeira, I. Shames, H. Sandberg, and K. H. Johansson, “A secure control framework for resource-limited adversaries,” Automatica, vol. 51, pp. 135–148, Jan. 2015. doi: 10.1016/j.automatica.2014.10.067
    [10]
    Q. R. Zhang, K. Liu, D. Y. Han, G. Z. Su, and Y. Q. Xia, “Design of stealthy deception attacks with partial system knowledge,” IEEE Trans. Autom. Control, vol. 68, no. 2, pp. 1069–1076, Feb. 2023. doi: 10.1109/TAC.2022.3146079
    [11]
    A. Teixeira, K. C. Sou, H. Sandberg, and K. H. Johansson, “Secure control systems: A quantitative risk management approach,” IEEE Control Syst. Mag., vol. 35, no. 1, pp. 24–45, Feb. 2015. doi: 10.1109/MCS.2014.2364709
    [12]
    H. S. Sánchez, D. Rotondo, T. Escobet, V. Puig, and J. Quevedo, “Bibliographical review on cyber attacks from a control oriented perspective,” Annu. Rev. Control, vol. 48, pp. 103–128, Sept. 2019. doi: 10.1016/j.arcontrol.2019.08.002
    [13]
    T. Y. Zhang and D. Ye, “False data injection attacks with complete stealthiness in cyber-physical systems: A self-generated approach,” Automatica, vol. 120, p. 109117, Oct. 2020. doi: 10.1016/j.automatica.2020.109117
    [14]
    K. K. Zhang, C. Keliris, T. Parisini, and M. M. Polycarpou, “Stealthy integrity attacks for a class of nonlinear cyber-physical systems,” IEEE Trans. Autom. Control, vol. 67, no. 12, pp. 6723–6730, Dec. 2022. doi: 10.1109/TAC.2021.3131656
    [15]
    A. Y. Lu and G. H. Yang, “Input-to-state stabilizing control for cyber-physical systems with multiple transmission channels under denial of service,” IEEE Trans. Autom. Control, vol. 63, no. 6, pp. 1813–1820, 2018. doi: 10.1109/TAC.2017.2751999
    [16]
    H. Zhang, P. Cheng, L. Shi, and J. M. Chen, “Optimal denial-of-service attack scheduling with energy constraint,” IEEE Trans. Autom. Control, vol. 60, no. 11, pp. 3023–3028, Nov. 2015. doi: 10.1109/TAC.2015.2409905
    [17]
    H. Zhang, Y. F. Qi, J. F. Wu, L. K. Fu, and L. D. He, “DoS attack energy management against remote state estimation,” IEEE Trans. Control Netw. Syst., vol. 5, no. 1, pp. 383–394, Mar. 2018. doi: 10.1109/TCNS.2016.2614099
    [18]
    S. Amin, A. A. Cárdenas, and S. S. Sastry, “Safe and secure networked control systems under denial-of-service attacks,” in Proc. 12th Int. Workshop on Hybrid Systems: Computation and Control, San Francisco, USA, 2009, pp. 31–45.
    [19]
    A. Teixeira, I. Shames, H. Sandberg, and K. H. Johansson, “Revealing stealthy attacks in control systems,” in Proc. 50th Annu. Allerton Conf. Communication, Control, and Computing, Monticello, USA, 2012, pp. 1806–1813.
    [20]
    F. Pasqualetti, F. Dörfler, and F. Bullo, “Attack detection and identification in cyber-physical systems,” IEEE Trans. Autom. Control, vol. 58, no. 11, pp. 2715–2729, Nov. 2013. doi: 10.1109/TAC.2013.2266831
    [21]
    R. S. Smith, “A decoupled feedback structure for covertly appropriating networked control systems,” IFAC Proc. Vol., vol. 44, no. 1, pp. 90–95, Jan. 2011. doi: 10.3182/20110828-6-IT-1002.01721
    [22]
    Y. B. Mao, H. Jafarnejadsani, P. Zhao, E. Akyol, and N. Hovakimyan, “Novel stealthy attack and defense strategies for networked control systems,” IEEE Trans. Autom. Control, vol. 65, no. 9, pp. 3847–3862, Sept. 2020. doi: 10.1109/TAC.2020.2997363
    [23]
    Y. L. Mo, R. Chabukswar, and B. Sinopoli, “Detecting integrity attacks on SCADA systems,” IEEE Trans. Control Syst. Technol., vol. 22, no. 4, pp. 1396–1407, Jul. 2014. doi: 10.1109/TCST.2013.2280899
    [24]
    R. M. G. Ferrari and A. M. H. Teixeira, “A switching multiplicative watermarking scheme for detection of stealthy cyber-attacks,” IEEE Trans. Autom. Control, vol. 66, no. 6, pp. 2558–2573, Jun. 2021. doi: 10.1109/TAC.2020.3013850
    [25]
    A. Hoehn and P. Zhang, “Detection of covert attacks and zero dynamics attacks in cyber-physical systems,” in Proc. American Control Conf., Boston, USA, 2016, pp. 302–307.
    [26]
    S. Weerakkody and B. Sinopoli, “Detecting integrity attacks on control systems using a moving target approach,” in Proc. 54th IEEE Conf. Decision and Control, Osaka, Japan, 2015, pp. 5820–5826.
    [27]
    P. Griffioen, S. Weerakkody, and B. Sinopoli, “A moving target defense for securing cyber-physical systems,” IEEE Trans. Autom. Control, vol. 66, no. 5, pp. 2016–2031, May 2021. doi: 10.1109/TAC.2020.3005686
    [28]
    M. M. Polycarpou and A. J. Helmicki, “Automated fault detection and accommodation: A learning systems approach,” IEEE Trans. Syst. Man Cybern., vol. 25, no. 11, pp. 1447–1458, Nov. 1995. doi: 10.1109/21.467710
    [29]
    S. X. Ding, Model-Based Fault Diagnosis Techniques: Design Schemes, Algorithms, and Tools. 2nd ed. London, UK: Springer, 2013.
    [30]
    M. Blanke, M. Kinnaert, J. Lunze, and M. Staroswiecki, Diagnosis and Fault-Tolerant Control. 2nd ed. Berlin, Germany: Springer, 2006.
    [31]
    Y. K. Wu, B. Jiang, and N. Y. Lu, “A descriptor system approach for estimation of incipient faults with application to high-speed railway traction devices,” IEEE Trans. Syst. Man Cybern. Syst., vol. 49, no. 10, pp. 2108–2118, Oct. 2019. doi: 10.1109/TSMC.2017.2757264
    [32]
    K. K. Zhang, B. Jiang, X. G. Yan, and Z. H. Mao, “Incipient fault detection for traction motors of high-speed railways using an interval sliding mode observer,” IEEE Trans. Intell. Transport. Syst., vol. 20, no. 7, pp. 2703–2714, Jul. 2019. doi: 10.1109/TITS.2018.2878909
    [33]
    C. Keliris, M. M. Polycarpou, and T. Parisini, “An integrated learning and filtering approach for fault diagnosis of a class of nonlinear dynamical systems,” IEEE Trans. Neural Netw. Learning Syst., vol. 28, no. 4, pp. 988–1004, Apr. 2017. doi: 10.1109/TNNLS.2015.2504418
    [34]
    M. Taheri, K. Khorasani, I. Shames, and N. Meskin, Cyber Attack and machine induced fault detection and isolation methodologies for cyber-physical systems, 2020. [Online]. Available: https://arxiv.org/abs/2009.06196.
    [35]
    K. K. Zhang, M. M. Polycarpou, and T. Parisini, “Enhanced anomaly detector for nonlinear cyber-physical systems against stealthy integrity attacks,” IFAC-PapersOnLine, vol. 53, no. 2, pp. 13682–13687, Jan. 2020. doi: 10.1016/j.ifacol.2020.12.870
    [36]
    K. K. Zhang, C. Keliris, M. M. Polycarpou, and T. Parisini, “Detecting stealthy integrity attacks in a class of nonlinear cyber-physical systems: A backward-in-time approach,” Automatica, vol. 141, p. 110262, Jul. 2022. doi: 10.1016/j.automatica.2022.110262
    [37]
    E. Kontouras, A. Tzes, and L. Dritsas, “Hybrid detection of intermittent cyber-attacks in networked power systems,” Energies, vol. 12, no. 24, p. 4625, Dec. 2019. doi: 10.3390/en12244625
    [38]
    S. Gao, H. Zhang, Z. P. Wang, C. Huang, and H. C. Yan, “Optimal injection attack strategy for cyber-physical systems under resource constraint: A game approach,” IEEE Trans. Control Netw. Syst., to be published.
    [39]
    J. Chen and R. J. Patton, Robust Model-Based Fault Diagnosis for Dynamic Systems. Springer, 2012.
    [40]
    X. D. Zhang, M. M. Polycarpou, and T. Parisini, “Fault diagnosis of a class of nonlinear uncertain systems with Lipschitz nonlinearities using adaptive estimation,” Automatica, vol. 46, no. 2, pp. 290–299, 2010. doi: 10.1016/j.automatica.2009.11.014
    [41]
    K. K. Zhang, B. Jiang, X. G. Yan, and J. Shen, “Interval sliding mode observer based incipient sensor fault detection with application to a traction device in China railway high-speed,” IEEE Trans. Veh. Technol., vol. 68, no. 3, pp. 2585–2597, 2019. doi: 10.1109/TVT.2019.2894670
    [42]
    M. Basseville and I. V. Nikiforov, Detection of Abrupt Changes: Theory and Application. Englewood Cliffs: Prentice-Hall, 1993.
    [43]
    B. D. Anderson and J. B. Moore, Optimal Filtering. North Chelmsford, USA: Courier Corporation, 2012.
    [44]
    M. Green and D. J. N. Limebeer, Linear Robust Control. New York, USA: Dover Publications, 2012.
    [45]
    D. Simon, Optimal State Estimation: Kalman, H, and Nonlinear Approaches. Hoboken, USA: John Wiley & Sons, 2006.
    [46]
    G. A. Einicke, Smoothing, Filtering and Prediction: Estimating the Past, Present and Future. Rijeka: IntechOpen, 2012.
    [47]
    X. B. Li and K. M. Zhou, “A time domain approach to robust fault detection of linear time-varying systems,” Automatica, vol. 45, no. 1, pp. 94–102, Jan. 2009. doi: 10.1016/j.automatica.2008.07.017
    [48]
    R. N. Banavar and J. L. Speyer, “A linear-quadratic game approach to estimation and smoothing,” in Proc. American Control Conf., Boston, USA, 1991, pp. 2818–2822.
    [49]
    H. Abou-Kandil, G. Freiling, V. Ionescu, and G. Jank, Matrix Riccati Equations in Control and Systems Theory. Birkhäuser Verlag, Basel, 2012.
    [50]
    G. Basile and G. Marro, Controlled and Conditioned Invariants in Linear System Theory. Englewood Cliffs: Prentice Hall, 1992.
    [51]
    Q. J. Xia, M. Rao, Y. Q. Ying, and X. M. Shen, “Adaptive fading Kalman filter with an application,” Automatica, vol. 30, no. 8, pp. 1333–1338, Aug. 1994. doi: 10.1016/0005-1098(94)90112-0
    [52]
    Y. J. Zhang, J. F. Zhang, X. K. Liu, and Z. Liu, “Quantized-output feedback model reference control of discrete-time linear systems,” Automatica, vol. 137, p. 110027, Mar. 2022. doi: 10.1016/j.automatica.2021.110027
    [53]
    J. Guo, Y. J. Zhang, J. F. Zhang, and X. K. Liu, “Finite quantized-output feedback tracking control of possibly non-minimum phase linear systems,” IEEE Control Syst. Lett., vol. 6, pp. 2407–2412, Mar. 2022. doi: 10.1109/LCSYS.2022.3159130
    [54]
    M. L. Lv, W. W. Yu, J. D. Cao, and S. Baldi, “A separation-based methodology to consensus tracking of switched high-order nonlinear multiagent systems,” IEEE Trans. Neural Netw. Learn. Syst., vol. 33, no. 10, pp. 5467–5479, Oct. 2022. doi: 10.1109/TNNLS.2021.3070824
    [55]
    M. L. Lv, B. De Schutter, C. Shi, and S. Baldi, “Logic-based distributed switching control for agents in power-chained form with multiple unknown control directions,” Automatica, vol. 137, p. 110143, Mar. 2022. doi: 10.1016/j.automatica.2021.110143
    [56]
    Y. Liu, D. Y. Yao, L. J. Wang, and S. J. Lu, “Distributed adaptive fixed-time robust platoon control for fully heterogeneous vehicles,” IEEE Trans. Syst. Man Cybern. Syst., vol. 53, no. 1, pp. 264–274, Jan. 2023. doi: 10.1109/TSMC.2022.3179444
  • Related Articles

    [1]Mengli Wei, Wenwu Yu, Duxin Chen, Mingyu Kang, Guang Cheng. Privacy Distributed Constrained Optimization Over Time-Varying Unbalanced Networks and Its Application in Federated Learning[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(2): 335-346. doi: 10.1109/JAS.2024.124869
    [2]Xuyang Wang, Dengxiu Yu, Xiaodi Li. Impulsive Consensus of MASs With Input Saturation and DoS Attacks[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(2): 414-424. doi: 10.1109/JAS.2024.124944
    [3]Shouyan Chen, Weitian He, Zhijia Zhao, Yun Feng, Zhijie Liu, Keum-Shik Hong. Adaptive Control of a Flexible Manipulator With Unknown Hysteresis and Intermittent Actuator Faults[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(1): 148-158. doi: 10.1109/JAS.2024.124653
    [4]Xiang Chen, Yujuan Wang, Yongduan Song. Unifying Fixed Time and Prescribed Time Control for Strict-Feedback Nonlinear Systems[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(2): 347-355. doi: 10.1109/JAS.2024.124401
    [5]Ruotian Liu, Yihui Hu, Agostino Marcello Mangini, Maria Pia Fanti. K-Corruption Intermittent Attacks for Violating the Codiagnosability[J]. IEEE/CAA Journal of Automatica Sinica, 2025, 12(1): 159-172. doi: 10.1109/JAS.2024.124680
    [6]Feisheng Yang, Jiaming Liu, Xiaohong Guan. Distributed Fixed-Time Optimal Energy Management for Microgrids Based on a Dynamic Event-Triggered Mechanism[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(12): 2396-2407. doi: 10.1109/JAS.2024.124686
    [7]Dan Zhang, Jiabin Hu, Jun Cheng, Zheng-Guang Wu, Huaicheng Yan. A Novel Disturbance Observer Based Fixed-Time Sliding Mode Control for Robotic Manipulators With Global Fast Convergence[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(3): 661-672. doi: 10.1109/JAS.2023.123948
    [8]Mahdi Taheri, Khashayar Khorasani, Nader Meskin. On Zero Dynamics and Controllable Cyber-Attacks in Cyber-Physical Systems and Dynamic Coding Schemes as Their Countermeasures[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(11): 2191-2203. doi: 10.1109/JAS.2024.124692
    [9]Xinli Shi, Xiangping Xu, Guanghui Wen, Jinde Cao. Fixed-Time Gradient Flows for Solving Constrained Optimization: A Unified Approach[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(8): 1849-1864. doi: 10.1109/JAS.2023.124089
    [10]Xiangmin Tan, Chunyan Hu, Guanzhen Cao, Qinglai Wei, Wei Li, Bo Han. Fixed-Time Antidisturbance Consensus Tracking for Nonlinear Multiagent Systems With Matching and Mismatching Disturbances[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(6): 1410-1423. doi: 10.1109/JAS.2024.124461
    [11]Jianquan Yang, Chunxi Yang, Xiufeng Zhang, Jing Na. Fixed-Time Sliding Mode Control With Varying Exponent Coefficient for Modular Reconfigurable Flight Arrays[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(2): 514-528. doi: 10.1109/JAS.2023.123645
    [12]Qiuyu Zhang, Lipeng Wang, Hao Meng, Wen Zhang, Genghua Huang. A LiDAR Point Clouds Dataset of Ships in a Maritime Environment[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(7): 1681-1694. doi: 10.1109/JAS.2024.124275
    [13]Qun Lu, Xiang Wu, Jinhua She, Fanghong Guo, Li Yu. Disturbance Rejection for Systems With Uncertainties Based on Fixed-Time Equivalent-Input-Disturbance Approach[J]. IEEE/CAA Journal of Automatica Sinica, 2024, 11(12): 2384-2395. doi: 10.1109/JAS.2024.124650
    [14]Dongdong Yue, Simone Baldi, Jinde Cao, Qi Li, Bart De Schutter. Distributed Adaptive Resource Allocation: An Uncertain Saddle-Point Dynamics Viewpoint[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(12): 2209-2221. doi: 10.1109/JAS.2023.123402
    [15]Yongbao Wu, Ziyuan Sun, Guangtao Ran, Lei Xue. Intermittent Control for Fixed-Time Synchronization of Coupled Networks[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(6): 1488-1490. doi: 10.1109/JAS.2023.123363
    [16]Jiayu Chai, Qiang Lu, Xudong Tao, Dongliang Peng, Botao Zhang. Dynamic Event-Triggered Fixed-Time Consensus Control and Its Applications to Magnetic Map Construction[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(10): 2000-2013. doi: 10.1109/JAS.2023.123444
    [17]Hongru Ren, Hui Ma, Hongyi Li, Zhenyou Wang. Adaptive Fixed-Time Control of Nonlinear MASs With Actuator Faults[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(5): 1252-1262. doi: 10.1109/JAS.2023.123558
    [18]Xiaohua Ge, Qing-Long Han, Qing Wu, Xian-Ming Zhang. Resilient and Safe Platooning Control of Connected Automated Vehicles Against Intermittent Denial-of-Service Attacks[J]. IEEE/CAA Journal of Automatica Sinica, 2023, 10(5): 1234-1251. doi: 10.1109/JAS.2022.105845
    [19]Ruizhuo Song, Liao Zhu. Optimal Fixed-Point Tracking Control for Discrete-Time Nonlinear Systems via ADP[J]. IEEE/CAA Journal of Automatica Sinica, 2019, 6(3): 657-666. doi: 10.1109/JAS.2019.1911453
    [20]Wenhui Liu, Feiqi Deng, Jiarong Liang, Haijun Liu. Distributed Average Consensus in Multi-agent Networks with Limited Bandwidth and Time-delays[J]. IEEE/CAA Journal of Automatica Sinica, 2014, 1(2): 193-203.

Catalog

    通讯作者: 陈斌, bchen63@163.com
    • 1. 

      沈阳化工大学材料科学与工程学院 沈阳 110142

    1. 本站搜索
    2. 百度学术搜索
    3. 万方数据库搜索
    4. CNKI搜索

    Figures(8)

    Article Metrics

    Article views (613) PDF downloads(108) Cited by()

    Highlights

    • A stealthy intermittent integrity attack generation strategy is formulated, which does not require that the adversary has precise knowledge of the system states. A backward-in-time detection residual is formulated, which increases in magnitude each time a new attack occurs
    • An optimal fixed-point smoother with covariance matrix resetting is proposed to implement the aforementioned backward-in-time residual. Such a smoother guarantees robustness to both disturbances and noise, and can also reset the covariance matrix to maintain sensitivity to intermittent integrity attacks
    • The corresponding adaptive threshold is designed, and an attack detectability analysis is carried out to characterize quantitatively the class of detectable stealthy intermittent integrity attacks

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return