Skip to main content

Theory and Modern Applications

Design of sampled-data controllers for the synchronization of complex dynamical networks under controller attacks


This paper addresses the design problem of sampled-data controllers for the synchronization of complex dynamical networks under controller attack. Some discontinuous Lyapunov functionals and zero equation are employed to deal with a sampled-data pattern. Due to the limited access of networks to hackers, cyber-attacks on the controller are considered randomly occurring and are described as an attack function which is nonlinear but assumed to be certain known conditions. The designing conditions for sampled-data controllers against cyber-attack are developed in terms of linear matrix inequality (LMI) by using Lyapunov theory and some novel inequalities. Finally, a numerical example is given to prove the usefulness of the proposed method.

1 Introduction

Nowadays, most of systems, components, plants are connected complicatedly to each other thanks to high level of communication technology, which can be called complex dynamic networks (CDNs). CDNs consist of many nodes with complicated interconnections among them, so each node is affected and affects the others whether they are willing or not; for example, World Wide Web (WWW), social media, metabolic systems, Internet of Things (IoT), transportation networks, power grids, food chain, and so on. By considering the high potential of CDNs, therefore, it is natural that the research on CDNs has become a large research area, and many researchers have discovered the nature of CDNs [1,2,3]. The synchronization is one of the popular applications of CDNs, which makes all states of each node in a CDN follow the same trajectories, and it has been widely studied [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]. To solve the synchronization problem for CDNs, various control schemes, such as feedback control [4,5,6], adaptive control [7, 8], sliding-mode control [9, 10], dynamic control [11, 12], event-triggered control [13], impulsive control [14, 15], \(\mathcal{H}_{\infty }\) control [16], and pinning control [17,18,19,20], have been employed.

In [17], the problem of both exponential synchronization and generalized synchronization of CDNs with/without time-varying coupling delays was investigated by designing an intermittent periodically adaptive pinning controller, and the results were applied to the nearest-neighbor network and the Barabasi–Albert network to show the effectiveness. The paper [7] proposed a concept of consecutive synchronization and designed a distributed adaptive controller for consecutive synchronization of CDNs with special topology structures. Synchronization of CDNs with N-coupled fractional-order chaotic system oscillators and regular/irregular topologies was studied in [6], and their results were applied to various fractional-order chaotic systems, Lorenz, Volta, Duffing, and financial chaotic oscillators. [8] dealt with traffic road network as an application of CDNs, in which an adaptive controller and adjustable coupling strength were designed to reduce the impact of uncertainties, and pinning control method was employed as well for the smooth flow of traffic.

Rapidly growing communication technology opened a research field on sampled-data control. Sampled-data control has many benefits such as easy installation and maintenance, low operation fee, and so on. Therefore, it is natural to have a meteoric rise of research field on the sampled-data control scheme [21,22,23,24,25,26,27,28,29,30,31,32]. In [21], an aperiodic sampled-data controller was designed for controlling the two-wheel inverted pendulum in the T-S fuzzy model with disturbances on actuators where the proposed controller satisfies very-strict passivity. In [22], \(\mathcal{H}_{\infty }\) state estimator for genetic regulatory networks with random delays, uncertainties, and disturbances was designed using sampled-data of the concentrations of mRNAs and proteins. In [23], an observer-based sampled-data controller was designed for a class of scalar nonlinear affine systems where the discrete-time states were firstly estimated by a nonlinear state observer, and then they were used for designing the sampled-data control. And there are a number of papers on the synchronization of CDNs by a sampled-data controller. The synchronization of CDNs with coupling time-varying delays was achieved via a sampled-data controller in [24], in which sampled-data had a constant time delay, and a discontinuous Lyapunov functional based on the extended Wirtinger inequality was used. In [25], the synchronizability analysis for CDNs with sampled-data coupling signals was conducted by employing multiple-integral Lyapunov functionals, and then a sampling period-dependent criterion was derived in terms of linear matrix inequalities (LMIs). The main feature of the sampled-data controller is that it uses discontinuous signals because the control signals can be updated at sampling instants and are held a constant value for sampling periods. To make the efficient use of this feature, several techniques have been reported, such as input delay approach [26, 27], discontinuous Lyapunov functional based on extended Wirtinger inequality [28], free-matrix-based time-dependent discontinuous Lyapunov approach [30], looped-functional-based approach [29], sampling-instant-to-present-time fragmentation approach [31, 32], and so on.

On the other hand, from the recent situation of growing communication technology, it can be understandably expected that the importance of security against cyber-attack has been emphasized [33,34,35,36,37,38,39,40,41,42]. Adversaries maliciously modified the data by exploiting the vulnerabilities of networks, which has degraded or destroyed the stability and performance of the system. Mainly two kinds of cyber-attacks exist [33, 34]: one is denial of service (DoS), another is deception attacks. DoS attacks attempt to block traffic from the actuator and sensor and as a result bring the absence of data for the related components; this situation is well known as the packet dropout. On the other hand, the data from components can be substituted secretly for the data that adversaries want by hackers which is called deception attacks. Thus, to develop criteria with prevention measure against cyber-attacks is a major concern of scholars [38, 39]. In [40], \(\mathcal{H}_{\infty }\) observer-based periodic event-triggered controller for the control of a class of cyber-physical systems was designed under consideration of DoS attacks where the designed controller maximized the frequency and duration of the DoS attacks. In [41], distributed event-triggered \(\mathcal{H}_{\infty }\) filters were designed on sensor networks in the presence of sensor saturations and randomly occurring cyber-attacks. In [42], the consensus problem of nonlinear multi-agent systems was studied by using sampled agents’ states and considering cyber-attacks on connectivity of the network topology, but it was recoverable. Authors in [43] proposed both event- and self-triggered control schemes for the leader-following consensus problem of multi-agent systems under consideration of DoS attacks, in which both synchronous and asynchronous updated strategies for control protocols were derived. Obviously, cyber-attacks would become a hotter issue for CDNs because CDNs consist of numerous nodes connected through communication networks. Nevertheless, research on CDNs with the consideration of cyber-attacks deserves much interest from scholars, only several works have been reported, which motivates this research.

This paper is about the synchronization of CDNs using sampled-data information under controller attacks. The main contribution of the paper is laid on mainly two streams: (i) Development of novel discontinuous Lyapunov functionals and zero inequality. Novel discontinuous Lyapunov functional would help the utilization of sampling characteristic. Combining with the novel discontinuous Lyapunov functionals, a zero inequality is needed to deal with a rest term of the time derivative of the novel discontinuous Lyapunov functionals. These may lead to less conservative conditions for designing a sampled-data controller. (ii) Consideration of cyber-attacks on controllers. Recently, the concept of cyber-attacks is the hottest issue in control engineering, and many works on cyber-attacks have been done. However, only few works cover controllers under cyber-attacks, even controllers of CDNs with cyber-attacks are few or nonexistent. Therefore, this paper considers deception attacks on controllers and designs the suitable controller against the attacks.

2 Problem formulation

Consider the following a CDN with N linearly coupled identical nodes:

$$\begin{aligned} \dot{x}_{i}(t)=Ax(t)+Bf\bigl(x_{i}(t)\bigr)+\sum _{j=1}^{N} c_{ij}x_{j}(t)+u_{i}(t), \quad i=1,\ldots,N, \end{aligned}$$

where \(x_{i}=(x_{i1}, x_{i2}, \ldots, x_{in})^{T} \in \mathbb{R}^{n}\) is the state vector of the ith node, \(A\in \mathbb{R}^{n\times n}\), \(B\in \mathbb{R}^{n\times n}\) are known constant matrices, \(f(t)=(f_{1}(t), f_{2}(t), \ldots, f_{n}(t))^{T} : \mathbb{R}^{n} \rightarrow \mathbb{R}^{n}\) is a smooth nonlinear vector field, and \(u_{i}(t)=(u_{i1}, u_{i2}, \ldots, u_{in})^{T}\) is the control input of ith node. \(C = (c_{ij})_{N \times N}\) is the coupling matrix of the network, where each element of the matrix \(c_{ij}\) is defined as follows: if there is a connection from node i to node j (\(i\neq j\)), then \(c_{ij}=1\); otherwise \(c_{ij}=0\) (\(i\neq j\)), and the diagonal elements of matrix C are assumed by

$$ c_{ii}=-\sum_{j=1,j\neq i}^{N} c_{ij}=-\sum_{j=1,j\neq i}^{N} c_{ji}, \quad i=1,\ldots,N. $$

Assumption 1

The nonlinear function \(f_{i}(\cdot)\) (\(i=1,2, \ldots,n\)) satisfies the following condition:

$$\begin{aligned} 0\leq \frac{f_{i}(a)-f_{i}(b)}{a-b}\leq l_{i}, \qquad f_{i}(0)=0, \end{aligned}$$

where \(l_{i}\) is a positive constant.

Let us consider a target note as \(\dot{s}(t)=As(t)+Bf(s(t))\). Then the aim of this paper is synchronizing all nodes of a CDN up to the target node, i.e., \(\lim_{t\rightarrow \infty }\|s(t)-x_{i}(t)\|=0\) for \(i=1,2,\ldots,N\). To this end, we define error vectors as \(e_{i}(t)=s(t)-x _{i}(t)\). Then the error dynamics is given as follows:

$$\begin{aligned} \dot{e}_{i}(t)= {}& Ae(t)+B\bar{f}_{i}(t)-\sum _{j=i}^{N} c_{ij}e_{j}(t)-u _{i}(t), \quad i=1,\ldots,N, \end{aligned}$$

where \(\bar{f}_{i}(t)=f(s(t))-f(x_{i}(t))\).

In this paper, the controllers are designed using sampled-data signals as follows:

$$\begin{aligned} u_{i}^{F}(t)=K_{i}e_{i}(t_{k}), \quad t_{k} \leq t < t_{k+1}, \end{aligned}$$

where \(K_{i}\) is the gain matrices for ith node to be determined, \(t_{k}\) is the updating instant time of the Zero-Order-Hold (ZOH) satisfying \(t_{k+1}-t_{k}=h\), h is a positive scalar.

Recently, the controller has become a target of hackers to alter the transmitted information. So, this paper is concerned with designing a sampled-data controller under cyber-attack. Because of the limited communication capacity of the network resources for hackers, the attacks might randomly destroy the control signals, and the maliciously modified signals would not be far away from the original one. To reflect this situation, we introduce Bernoulli random variable \(\gamma _{i}(t) \in \{0, 1\}\), satisfying \(\Pr \{\gamma _{i}(t)=1\}=\gamma _{i}\), where \(\Pr \{x\}\) implies the occurrence probability of the event x, and \(\gamma _{i}\) is a known positive constant less than 1. Here, \(\gamma _{i}(t)=1\) means the cyber-attacks are launched to the controller \(u_{i}^{F}(t)\). The specific controller model under cyber-attack is as follows:

$$\begin{aligned} u_{i}(t) = {}& \bigl(1-\gamma _{i}(t)\bigr)K_{i}e_{i}(t_{k})+ \gamma _{i}(t)K_{i}g _{i}\bigl(e_{i} \bigl(t-d(t)\bigr)\bigr), \end{aligned}$$

where \(g_{i}(t)=(g_{i1}(t), g_{i2}(t), \ldots, g_{in}(t))^{T} : \mathbb{R}^{n} \rightarrow \mathbb{R}^{n}\) is the function of controller attacks, \(d(t)\) is time-varying delay satisfying \(0\leq d(t)\leq d\) and \(\dot{d}(t)\leq \mu \), and d and μ are known positive constants.

Assumption 2

The cyber-attack function \(g_{i}(\cdot)\) is satisfied \(g_{i}(0)=0\) and \(\forall i\in \{1,2,\ldots,n\}\), \(a\neq b\)

$$\begin{aligned} 0\leq \frac{g_{ij}(a)-g_{ij}(b)}{a-b}\leq m_{ij}, \end{aligned}$$

where \(m_{ij}\) is a positive constant.

Remark 1

It can be strictly said that the cyber-attacks considered in this paper are randomly occurring deception attacks because, when a controller is under cyber-attacks, the real control inputs are replaced with the incorrect ones which adversaries want. And the cyber-attacks would be launched at any time, this introduces randomly occurring sense. Since adversaries try to adjust real control signals and transmit them through communication networks, the considered cyber-attacks signals contain time-delayed information.

Then the closed-loop error systems with controller (4) can be rewritten in the following vector-matrix form:

$$\begin{aligned} \dot{e}(t)={} &A_{N}e(t)+B_{N}F\bigl(e(t) \bigr)-C_{N}e(t)-\bar{\varGamma }(t)Ke(t_{k}) \\ &{}-\varGamma (t) G\bigl(e\bigl(t-d(t)\bigr)\bigr), \end{aligned}$$

where \(e(t)=[e_{1}^{T}(t),e_{2}^{T}(t),\ldots ,e_{N}^{T}(t)]^{T}\), \(F(e(t))=[\bar{f}_{1}^{T}(e_{1}(t)),\bar{f} _{2}^{T}(e_{2}(t)),\ldots ,\bar{f}_{N}^{T}(e_{N}(t))]^{T}\), \(G(e(t))=[g_{1}^{T}(e_{1}(t)),g_{2}^{T}(e_{2}(t)),\ldots ,g_{N}^{T}(e_{N}(t))]^{T}\), \(K=\operatorname{diag}\{K_{1},K_{2},\ldots ,K_{N}\}\), \(\varGamma (t)=\operatorname{diag}\{\gamma _{1}(t),\gamma _{2}(t),\ldots , \gamma _{N}(t)\}\otimes I_{n}\), \(\bar{\varGamma }(t)= (I_{N}-\varGamma (t) )\otimes I_{n}\), \(A_{N}=I_{N}\otimes A\), \(B_{N}=I_{N}\otimes B\), \(C_{N}=C\otimes I_{n}\), and stands for the notation of Kronecker product.

3 Main results

This section proposes a design scheme for the sampled-data controller under cyber-attack synchronizing a CDN. Before proceeding further, the following lemma is given.

Lemma 1

For given three scalars a, b, c where \(a\le b\le c\), a positive definite matrix \(W\in \mathbb{R}^{n\times n}\), and any matrix \(S\in \mathbb{R}^{2n\times 2n}\) and all continuous function η in \([a,c]\rightarrow \mathbb{R}^{n}\) the following inequality holds:

$$\begin{aligned} \int _{a}^{c} \dot{\eta }^{T}(s)W\dot{\eta }(s)\,ds\geq \beta ^{T}(t) \mathcal{W}\beta (t), \end{aligned}$$

subject to

[ diag { W , 3 W } S diag { W , 3 W } ] >0,


$$\begin{aligned} &\beta (t)= \bigl[\varOmega _{1}^{T}(a,b), \varOmega _{2}^{T}(a,b), \varOmega _{1} ^{T}(b,c), \varOmega _{2}^{T}(b,c) \bigr]^{T}, \\ &\varOmega _{1}(a,b)=\eta (b)-\eta (a), \\ &\varOmega _{2}(a,b)=\eta (b)+\eta (a)-\frac{2}{b-a} \int _{a}^{b}\eta (s)\,ds. \end{aligned}$$


It is easy to derive Lemma 1 when we combined with Wirtinger-based integral inequality [44] and reciprocal convex lemma [45].

The following notations are used in the paper. The block entry matrices and a vector are defined as \(b_{i}\) \((i=1,\ldots,10)\in \mathbb{R} ^{10nN\times nN}\) (for example, \(b_{3}=[0,0,I,\underbrace{0, \ldots,0}_{7}]\)) and

$$\begin{aligned} \zeta (t)={} & \biggl[e^{T}(t), e^{T}\bigl(t-d(t)\bigr), e^{T}(t-d), \frac{1}{d(t)} \int _{t-d(t)}^{t}e^{T}(s)\,ds, \\ &{}\frac{1}{d-d(t)} \int _{t-d}^{t-d(t)}e^{T}(s)\,ds, e^{T}(t_{k}), \int _{t_{k}}^{t} e^{T}(s)\,ds, \dot{e}^{T}(t), \\ &{}F^{T}(t), G^{T}\bigl(t-d(t)\bigr) \biggr]^{T}. \end{aligned}$$


Now, the main result is given by the following theorem.

Theorem 1

For given scalars d, h, μ, α, \(\gamma _{i}\) (\(i=1,2,\ldots,N\)), \(l_{i}\) (\(i=1,2,\ldots,n\)), \(m_{ij}\) (\(i=1,2,\ldots,N\); \(j=1,2,\ldots,n\)), there exists a sampled-data controller with cyber-attack (4) for the synchronization of CDN (1) if there exist positive-definite matrices \(P, Q_{1}, Q_{2}, Q_{3}\in \mathbb{R}^{nN\times nN}\),

$$R=\left [ \textstyle\begin{array}{c|c} R_{1} & \textstyle\begin{array}{cc} R_{2} & R_{3} \end{array}\displaystyle \\ \hline \textstyle\begin{array}{c} \star \\ \star \end{array}\displaystyle & R_{4} \end{array}\displaystyle \right ]\in \mathbb{R}^{3nN\times 3nN}, $$

positive diagonal matrices \(U_{1}, U_{2}, X_{1}, X_{2}\in \mathbb{R}^{nN\times nN}\), a symmetric matrix \(H_{5}\in \mathbb{R}^{nN\times nN}\), a diagonal matrix \(D\in \mathbb{R}^{nN\times nN}\), any matrices \(H_{1}, H_{2}, H_{3}, H _{4}, \bar{K}\in \mathbb{R}^{nN\times nN}\), \(S\in \mathbb{R}^{2nN \times 2nN}\), \(Y\in \mathbb{R}^{10nN\times nN}\) satisfying the following LMIs:

$$\begin{aligned} &\mathcal{P}=\operatorname{diag}\{P,0,0\}+hH >0, \end{aligned}$$
Q= [ diag { Q 3 , 3 Q 3 } S diag { Q 3 , 3 Q 3 } ] >0,
$$\begin{aligned} &R_{4} >0, \end{aligned}$$
$$\begin{aligned} &\varUpsilon _{1}+h\varUpsilon _{2} < 0, \end{aligned}$$
[ ϒ 1 h b 6 R 1 b 6 T Y 1 h R 4 ] <0,


ϒ 1 = Sym { b 1 P b 8 T b 6 R 2 ( b 1 b 6 ) T b 6 R 3 b 7 T + Y Π 5 T + b 1 L U 1 b 9 T ϒ 1 = + b 2 M X 1 b 10 T + ( b 1 + α b 8 ) ( D b 8 T + D ( A N C N ) b 1 T + D B N b 9 T Γ ¯ K ¯ b 6 T ϒ 1 = Γ K ¯ b 10 T ) } Π 1 H Π 1 T + b 1 ( Q 1 + Q 2 ) b 1 T ( 1 μ ) b 2 Q 1 b 2 T b 3 Q 2 b 3 T ϒ 1 = + d b 8 Q 3 b 8 T Π 3 Q Π 3 T + b 1 L U 2 L b 1 T b 9 ( 2 U 1 + U 2 ) b 9 T + b 2 M X 2 M b 2 T ϒ 1 = b 10 ( 2 X 1 + X 2 ) b 10 T , ϒ 2 = Sym { Π 1 H Π 2 T } + Π 4 R Π 4 T , H = [ H 1 + H 1 T H 1 H 2 H 3 H 2 + H 2 T H 4 H 5 ] , Π 1 = [ b 1 , b 6 , b 7 ] , Π 2 = [ b 8 , 0 , b 1 b 6 ] , Π 3 = [ b 1 b 2 , b 1 + b 2 2 b 4 , b 2 b 3 , b 2 + b 3 2 b 5 ] , Π 4 = [ b 6 , b 8 , b 1 ] , Π 5 = [ b 1 b 6 , b 7 ] , L = I N diag { l 1 , l 2 , , l n } , M = diag { M 1 , M 2 , , M N } , M i = diag { m i 1 , m i 2 , , m i n } , Γ = diag { γ 1 , γ 2 , , γ N } I n , Γ ¯ = ( I N Γ ) I n ,

with the notation \(\operatorname{Sym}\{X\}\) indicating \(X+X^{T}\). Also, the desired control gain matrices (4) can be given by \(K=D^{-1}\bar{K}\).


Consider the following discontinuous Lyapunov functional for the error system (5):

$$\begin{aligned} V(t)=V_{1}(t)+V_{2}(t)+V_{3}(t), \quad t\in [t_{k}, t_{k+1}), \end{aligned}$$


$$\begin{aligned} &V_{1}(t)= e^{T}(t)Pe(t)+(t_{k+1}-t)\nu _{1}^{T}(t)H\nu _{1}(t), \\ &V_{2}(t)= \int _{t-d(t)}^{t}e^{T}(s)Q_{1}e(s)\,ds+ \int _{t-d}^{t}e^{T}(s)Q _{2}e(s)\,ds \\ &\hphantom{V_{2}(t)=}{}+ \int _{-d}^{0} \int _{t+\theta }^{t}\dot{e}^{T}(s)Q_{3} \dot{e}(s)\,ds\,d \theta, \\ &V_{3}(t)= (t_{k+1}-t) \int _{t_{k}}^{t}\nu _{2}^{T}(s)R \nu _{2}(s)\,ds, \end{aligned}$$


ν 1 ( t ) = [ e T ( t ) , e T ( t k ) , t k t e T ( s ) d s ] T , ν 2 ( t ) = [ e T ( t k ) , e ˙ T ( t ) , e T ( t ) ] T .

The matrix H is not defined as positive definite, so we firstly show the positiveness of \(V_{1}(t)\).

$$\begin{aligned} V_{1}(t)={} & \biggl(\frac{t_{k+1}-t}{h}+\frac{t-t_{k}}{h} \biggr)e^{T}(t)Pe(t)+(t _{k+1}-t)\nu _{1}^{T}(t)H \nu _{1}(t) \\ ={} &\frac{t_{k+1}-t}{h}\nu _{1}^{T}(t)\mathcal{P}\nu _{1}(t)+\frac{t-t _{k}}{h}e^{T}(t)Pe(t). \end{aligned}$$

From LMI (6), we can know \(V_{1}(t)\) is positive definite. And next, it is obvious that (i) \(V_{3}(t)\) is discontinuous at \(t=t_{k}\), i.e., \(V_{3}(t_{k})=\lim_{t\rightarrow t_{k}^{+}}V_{3}(t)\neq \lim_{t\rightarrow t_{k}^{-}}V_{3}(t)\); (ii) \(V_{3}(t)\) is positive during \(t\in (t_{k-1},t_{k})\); (iii) \(V_{3}(t_{k})=0\). Therefore, we can conclude \(\lim_{t\rightarrow t_{k}^{-}}V(t)\geq V(t_{k})\).

The infinitesimal operator \(\mathcal{L}\) of \(V(x_{t})\) is defined as follows:

$$\begin{aligned} \mathcal{L}V(x_{t})=\lim_{c\rightarrow 0^{+}}\frac{1}{c} \bigl\{ \mathbb{E} \bigl\{ V(x_{t+h})|x_{t}\bigr\} \bigr\} , \end{aligned}$$

where \(\mathbb{E}\{x\}\) means the expectation of the stochastic variable x.

Then we can have

$$\begin{aligned} \mathcal{L}V_{1}(t)={} &2e^{T}(t)P\dot{e}(t)+2(t_{k+1}-t) \nu _{1}^{T}(t)H \nu _{3}(t)-\nu _{1}^{T}(t)H\nu _{1}(t) \\ ={} &\zeta ^{T}(t) \bigl(\operatorname{Sym}\bigl\{ b_{1}Pb_{8}^{T}+(t_{k+1}-t) \varPi _{1} H \varPi _{2} ^{T}\bigr\} -\varPi _{1}H\varPi _{1}^{T} \bigr)\zeta (t), \end{aligned}$$

where \(\varPi _{1}\) and \(\varPi _{2}\) are defined in Theorem 1.

Also, we have

$$\begin{aligned} \mathcal{L}V_{2}(t)={} &e^{T}(t) (Q_{1}+Q_{2})e(t)- \bigl(1-\dot{h(t)}\bigr)e^{T}\bigl(t-d(t)\bigr)Q _{1}e \bigl(t-d(t)\bigr) \\ &{}-e^{T}(t-d)Q_{2}e(t-d)+d\dot{e}^{T}(t)Q_{3} \dot{e}(t)- \int _{t-d} ^{t}\dot{e}^{T}(s)Q_{3} \dot{e}(s)\,ds \\ \leq{} &\zeta ^{T}(t) \bigl(b_{1}(Q_{1}+Q_{2})b_{1}^{T}-(1- \mu)b_{2}Q _{1}b_{2}^{T}-b_{3}Q_{2}b_{3}^{T}+db_{8}Q_{3}b_{8}^{T} \bigr)\zeta (t) \\ &{}- \int _{t-d}^{t}\dot{e}^{T}(s)Q_{3} \dot{e}(s)\,ds \\ \leq{} &\zeta ^{T}(t) \bigl(b_{1}(Q_{1}+Q_{2})b_{1}^{T}-(1- \mu)b_{2}Q _{1}b_{2}^{T}-b_{3}Q_{2}b_{3}^{T}+db_{8}Q_{3}b_{8}^{T} \\ &{}-\varPi _{3}\mathcal{Q}\varPi _{3}^{T} \bigr) \zeta (t), \end{aligned}$$

where \(\varPi _{3}\) and \(\mathcal{Q}\) are defined in Theorem 1, and Lemma 1 with LMI (7) is used to obtain the last equation.


$$\begin{aligned} \mathcal{L}V_{3}(t)={} &(t_{k+1}-t)\nu _{2}^{T}(t)R \nu _{2}(t)- \int _{t_{k}} ^{t}\nu _{2}^{T}(s)R \nu _{2}(s)\,ds \\ ={} &(t_{k+1}-t)\nu _{2}^{T}(t)R\nu _{2}(t)-(t-t_{k})e^{T}(t_{k})R_{1}e(t _{k}) \\ &{}-2e^{T}(t_{k})R_{2}\bigl(e(t)-e(t_{k}) \bigr)-2e^{T}(t_{k})R_{3} \int _{t_{k}} ^{t}e(s)\,ds \\ &{}- \int _{t_{k}}^{t}\nu _{4}^{T}(s)R_{4} \nu _{4}(s)\,ds \\ ={} &\zeta ^{T}(t) \bigl((t_{k+1}-t)\varPi _{4}R \varPi _{4}^{T}-(t-t_{k})b_{6}R _{1}b_{6}^{T} \\ &{}-\operatorname{Sym} \bigl\{ b_{6}R_{2}(b_{1}-b_{6})^{T}+b_{6}R_{3}b_{7}^{T} \bigr\} \bigr)\zeta (t) \\ &{}- \int _{t_{k}}^{t}\nu _{4}^{T}(s)R_{4} \nu _{4}(s)\,ds, \end{aligned}$$

where \(\varPi _{4}\) is defined in Theorem 1 and

ν 4 (t)= [ e ˙ T ( t ) , e T ( t ) ] T .

If condition (8) holds, then we have

0 = 2 ζ T ( t ) Y ( [ e ( t ) e ( t k ) t k t e ( s ) d s ] t k t ν 4 ( s ) d s ) ( t t k ) ζ T ( t ) Y R 4 1 Y T ζ ( t ) + 2 ζ T ( t ) Y [ e ( t ) e ( t k ) t k t e ( s ) d s ] + t k t ν 4 T ( s ) d s R 4 ( t t k ) t k t ν 4 ( s ) d s ( t t k ) ζ T ( t ) Y R 4 1 Y T ζ ( t ) + 2 ζ T ( t ) Y [ e ( t ) e ( t k ) t k t e ( s ) d s ] + t k t ν 4 T ( s ) R 4 ν 4 ( s ) d s ,

where Jensen’s inequality [46] is used to obtain the last equation.

Therefore, the following inequality can be derived:

t k t ν 4 T ( s ) R 4 ν 4 ( s ) d s 2 ζ T ( t ) Y [ e ( t ) e ( t k ) t k t e ( s ) d s ] + ( t t k ) ζ T ( t ) Y R 4 1 Y T ζ ( t ) = ζ T ( t ) ( Sym { Y Π 5 T } + ( t t k ) Y R 4 1 Y T ) ζ ( t ) ,

where \(\varPi _{5}\) is defined in Theorem 1.

According to Assumptions 1 and 2, for positive diagonal matrices \(U_{1}\), \(U_{2}\), \(X_{1}\), \(X_{2}\), we can obtain the following inequalities:

$$\begin{aligned} 0\leq{} &2 \bigl(e^{T}(t)L-F^{T}\bigl(e(t)\bigr) \bigr)U_{1}F\bigl(e(t)\bigr)+e^{T}(t)LU_{2}Le(t)-F ^{T}\bigl(e(t)\bigr)U_{2}F\bigl(e(t)\bigr) \\ ={} &\zeta ^{T}(t) \bigl(b_{1}LU_{2}Lb_{1}^{T}-b_{9}(2U_{1}+U_{2}) \bigr)b_{9} ^{T}+\operatorname{Sym}\bigl\{ b_{1}LU_{1}b_{9}^{T} \bigr\} \big)\zeta (t), \end{aligned}$$
$$\begin{aligned} 0\leq{} &2 \bigl(e^{T}\bigl(t-d(t)\bigr)M-G^{T}\bigl(e \bigl(t-d(t)\bigr)\bigr) \bigr)X_{1}G\bigl(e\bigl(t-d(t)\bigr)\bigr) \\ &{}+e^{T}\bigl(t-d(t)\bigr)MX_{2}Me\bigl(t-d(t) \bigr)-G^{T}\bigl(e\bigl(t-d(t)\bigr)\bigr)X_{2}G\bigl(e \bigl(t-d(t)\bigr)\bigr) \\ ={} &\zeta ^{T}(t) \bigl(b_{2}MX_{2}Mb_{2}^{T}-b_{10}(2X_{1}+X_{2}) \bigr)b_{10} ^{T}+\operatorname{Sym}\bigl\{ b_{2}MX_{1}b_{10}^{T} \bigr\} \big)\zeta (t). \end{aligned}$$

Also, according to the error system (5), for a scalar α and a diagonal matrix D, the following equation holds:

$$\begin{aligned} 0={} &\mathbb{E} \bigl\{ 2 \bigl[e^{T}(t)D+\alpha \dot{e}^{T}(t)D \bigr] \bigl[-\dot{e}(t)+(A_{N}-C_{N})e(t)+B_{N}F(t) \\ &{}-\bar{\varGamma }Ke(t_{k})-\varGamma KG\bigl(t-d(t)\bigr) \bigr] \bigr\} \\ ={} &\mathbb{E} \bigl\{ \zeta ^{T}(t) \bigl(\operatorname{Sym} \bigl\{ (b_{1}+\alpha b_{8}) \bigl(-Db _{8}^{T}+D(A_{N}-C_{N})b_{1}^{T}+DB_{N}b_{9}^{T} \\ &{}-\bar{\varGamma }\bar{K}b_{6}^{T}-\varGamma \bar{K}b_{10}^{T}\bigr) \bigr\} \bigr)\zeta (t) \bigr\} , \end{aligned}$$

where \(\bar{K}=DK\).

By adding Eqs. (15)–(18) to \(\mathbb{E}\{\mathcal{L}V(t) \}\), the new upper bound of \(\mathbb{E}\{\mathcal{L}V(t)\}\) can be obtained:

$$\begin{aligned} \mathbb{E}\bigl\{ \mathcal{L}V(t)\bigr\} \leq \mathbb{E} \biggl\{ \zeta ^{T}(t) \biggl(\frac{t_{k+1}-t}{h}(\varUpsilon _{1}+h \varUpsilon _{2})+ \frac{t-t_{k}}{h}(\varUpsilon _{1}+h \varUpsilon _{3}) \biggr) \zeta (t) \biggr\} , \end{aligned}$$


$$\begin{aligned} \varUpsilon _{3}= {}&YR_{4}^{-1}Y^{T}-b_{6}R_{1}b_{6}^{T}. \end{aligned}$$

By Schur complement, it is clear that LMIs (9) and (10) and \(\mathbb{E}\{\mathcal{L}V(t)\}<0\) are equivalent. In other words, the designed controller (4) makes the error system asymptotically stable against controller attacks. This completes the proof. □

4 Numerical examples

This section considers CDNs (1) consisting of five Chua’s chaotic circuits in which the parameters are given as follows:

A = [ a m 1 a 0 1 1 1 0 b 0 ] , B = [ a ( m 0 m 1 ) 0 0 0 0 0 0 0 0 ] , C = 0.2 × [ 3 1 1 0 1 1 4 1 1 1 1 1 3 1 0 0 1 1 3 1 1 1 0 1 3 ] , f j ( x i j ( t ) ) = 1 2 ( | x i j ( t ) + c | | x i j ( t ) c | ) , j = 1 , , n ; i = 1 , , N , a = 9 , b = 14.28 , c = 1 , m 0 = 1 / 7 , m 1 = 2 / 7 , l i = 1 ( i = 1 , , n ) .

For the simulation, we choose the following parameters and initial conditions: \(d(t)=0.4+0.1\sin (t)\), \(d=0.5\), \(h=0.4\), \(\alpha =0.1\), \(\gamma _{ij}=0.5\) (\(i=1,\ldots,N\); \(j=1,\ldots,n\)), \(g_{ij}(a)= \tanh (0.04a)\) (\(i=1,\ldots,N\); \(j=1,\ldots,n\)), \(x_{1}(0)=[-0.1\ {-}0.5\ {-}0.7]\), \(x_{2}(0)=[-0.1\ {-}0.4\ 0.3]\), \(x_{3}(0)=[0.6\ {-}1.5\ 0]\), \(x_{4}(0)=[0.1 \ 0.1\ 0.1]\), \(x_{5}(0)=[0\ 0.5\ {-}0.4]\), and \(s(0)=[0.1\ 0.5 \ {-}0.7]\).

With the above parameters, Theorem 1 calculates the following control gains:

$$\begin{aligned} K=\operatorname{diag}\{K_{1}, \ldots, K_{5}\}, \end{aligned}$$


K 1 = [ 4.8871 0.0018 0.0005 0.0016 0.6933 0.0005 0.0004 0.0010 4.8007 ] , K 2 = [ 4.8307 0.0011 0.0001 0.0011 0.6290 0.0000 0.0003 0.0002 4.7882 ] , K 3 = [ 4.8360 0.0005 0.0005 0.0000 0.6312 0.0006 0.0009 0.0002 4.8005 ] , K 4 = [ 4.8643 0.0026 0.0004 0.0019 0.6111 0.0007 0.0001 0.0012 4.8000 ] , K 5 = [ 4.8733 0.0014 0.0005 0.0008 0.6719 0.0023 0.0069 0.0030 4.8101 ] .

Then the controlled error signals can be obtained in Fig. 1 which shows the error dynamics asymptotically converse to zero; in other words, the synchronization is achieved between \(s(t)\) and \(x_{i}(t)\) for \(i=1,\ldots,N\). Figures 2 and 3 show the applied control signals \(u_{i}(t)\) and cyber-attack scenario \(\gamma _{i}(t)\), respectively. Figure 4 displays the attack signals transmitted to the CDNs instead of the real control signals when cyber-attacks are launched to the controller of ith node, i.e., \(\gamma _{i}(t)=1\). The difference between the real control signals(without cyber-attack) \(u_{i}^{F}(t)\) and the control signals under attack \(u_{i}(t)\) is depicted in Fig. 5. As seen in Figs. 1 and 5, the controller attacks make big changes from real control signals, but our designed controller works very well to synchronize all the states of CDNs up to the target.

Figure 1
figure 1

The error trajectories with controller (4) under cyber-attacks \(e(t)\)

Figure 2
figure 2

The applied control signals for each node under cyber-attacks \(u_{i}(t)\)

Figure 3
figure 3

The random cyber-attack scenario for each node \(\gamma _{i}(t)\)

Figure 4
figure 4

The attack signals to each node \(K_{i}g_{i}(e_{i}(t-d(t))\)

Figure 5
figure 5

The difference between control signals with/without cyber-attacks, \(u_{i}(t)-u_{i}^{F}(t)\)

5 Conclusions

In this paper, the design problem of sampled-data controllers for the synchronization of CDNs was investigated. To reflect a real-world situation, we considered the controllers under cyber-attacks which are randomly occurring. The maliciously reformed control data by adversaries was assumed to be a nonlinear function including time-varying delayed control information. Using discontinuous Lyapunov functionals and a zero inequality, the designing conditions of the sampled-data controller were derived in terms of LMI. The validity of the proposed method was proven by a numerical example.


  1. Strogatz, S.H.: Exploring complex networks. Nature 410, 268–276 (2001)

    Article  Google Scholar 

  2. Dorogovtsev, S.N., Mendes, J.F.F.: Evolution of networks. Adv. Phys. 51, 1079–1187 (2002)

    Article  Google Scholar 

  3. Newman, M.E.J.: The structure and function of complex networks. SIAM Rev. 45, 167–256 (2003)

    Article  MathSciNet  Google Scholar 

  4. Zhang, H., Cao, J., Xiong, L.: Novel synchronization conditions for time-varying delayed Lur’e system with parametric uncertainty. Appl. Math. Comput. 350, 224–236 (2019)

    Article  MathSciNet  Google Scholar 

  5. Wang, X., Gao, K., Chen, G., Xu, Y.: Finite-time synchronization control relationship analysis for two classes of Markovian jump complex networks under feedback control. Adv. Differ. Equ. 2018, 382 (2018)

    Article  MathSciNet  Google Scholar 

  6. Angulo-Guzman, S., Posadas-Castillo, C., Platas-Garza, M.A., Diaz-Romero, D.A., Lopez-Mancilla, D.: Chaotic synchronization of regular and irregular complex networks with fractional order oscillators. Int. J. Control. Autom. Syst. 14, 1114–1123 (2016)

    Article  Google Scholar 

  7. Kazemy, A., Cao, J.: Consecutive synchronization of a delayed complex dynamical network via distributed adaptive control approach. Int. J. Control. Autom. Syst. (2018).

    Article  Google Scholar 

  8. Wang, D., Che, W.-W., Yu, H., Li, J.-Y.: Adaptive pinning synchronization of complex networks with negative weights and its application in traffic road network. Int. J. Control. Autom. Syst. 16, 782–790 (2018)

    Article  Google Scholar 

  9. Ma, N., Liu, Z., Chen, L.: Sliding-mode \(\mathcal{H}_{\infty }\) synchronization for complex dynamical network systems with Markovian jump parameters and time-varying delays. Adv. Differ. Equ. 2019, 48 (2019)

    Article  MathSciNet  Google Scholar 

  10. Ali, M.S., Yogambigai, J., Cao, J.: Synchronization of master-slave Markovian switching complex dynamical networks with time-varying delays in nonlinear function via sliding mode control. Acta Math. Sci. 37, 368–384 (2017)

    Article  MathSciNet  Google Scholar 

  11. Lee, T.H., Park, J.H., Ji, D.H., Kwon, O.M., Lee, S.M.: Guaranteed cost synchronization of a complex dynamical network via dynamic feedback control. Appl. Math. Comput. 218, 6469–6481 (2012)

    MathSciNet  MATH  Google Scholar 

  12. Lee, T.H., Ji, D.H., Park, J.H., Jung, H.Y.: Decentralized guaranteed cost dynamic control for synchronization of a complex dynamical network with randomly switching topology. Appl. Math. Comput. 219, 996–1010 (2012)

    MathSciNet  MATH  Google Scholar 

  13. Sivaranjani, K., Rakkiyappan, R., Cao, J., Alsaedi, A.: Synchronization of nonlinear singularly perturbed complex networks with uncertain inner coupling via event triggered control. Appl. Math. Comput. 311, 283–299 (2017)

    MathSciNet  Google Scholar 

  14. Dai, A., Zhou, W., Feng, J., Fang, J., Xu, S.: Exponential synchronization of the coupling delayed switching complex dynamical networks via impulsive control. Adv. Differ. Equ. 2013, 195 (2013)

    Article  MathSciNet  Google Scholar 

  15. Lu, J., Ho, D.W.C., Cao, J., Kurths, J.: Single impulsive controller for globally exponential synchronization of dynamical networks. Nonlinear Anal., Real World Appl. 14, 581–593 (2013)

    Article  MathSciNet  Google Scholar 

  16. Lee, T.H., Park, J.H., Wu, Z.G., Lee, S.C., Lee, D.H.: Robust \(\mathcal{H}_{\infty }\) decentralized dynamic control for synchronization of a complex dynamical network with randomly occurring uncertainties. Nonlinear Dyn. 70, 559–570 (2012)

    Article  MathSciNet  Google Scholar 

  17. Pan, L., Tang, X., Pan, Y.: Generalized and exponential synchronization for a class of novel complex dynamic networks with hybrid time-varying delay via IPAPC. Int. J. Control. Autom. Syst. (2018).

    Article  Google Scholar 

  18. Shi, L., Chen, G., Zhong, S.: Outer synchronization of a class of mixed delayed complex networks based on pinning control. Adv. Differ. Equ. 2018, 330 (2018)

    Article  MathSciNet  Google Scholar 

  19. Li, B., Wang, N., Ruan, X., Pan, Q.: Pinning and adaptive synchronization of fractional-order complex dynamical networks with and without time-varying delay. Adv. Differ. Equ. 2018, 6 (2018)

    Article  MathSciNet  Google Scholar 

  20. Hu, J., Liang, J., Cao, J.: Synchronization of hybrid-coupled heterogeneous networks: pinning control and impulsive control schemes. J. Franklin Inst. 351, 2600–2622 (2014)

    Article  MathSciNet  Google Scholar 

  21. Liu, R., Wu, J., Wang, D.: Sampled-data fuzzy control of two-wheel inverted pendulums based on passivity theory. Int. J. Control. Autom. Syst. (2018).

    Article  Google Scholar 

  22. Chen, W., Chen, D., Hu, J., Liang, J., Dobaie, A.M.: A sampled-data approach to robust \(\mathcal{H}_{\infty }\) state estimation for genetic regulatory networks with random delays. Int. J. Control. Autom. Syst. 16, 491–504 (2018)

    Article  Google Scholar 

  23. Nguyen-Van, T.: An observer based sampled-data control for class of scalar nonlinear systems using continualized discretization method. Int. J. Control. Autom. Syst. 16, 709–716 (2018)

    Article  Google Scholar 

  24. Lee, T.H., Wu, Z.-G., Park, J.H.: Synchronization of a complex dynamical network with coupling time-varying delays via sampled-data control. Appl. Math. Comput. 219, 1354–1366 (2012)

    MathSciNet  MATH  Google Scholar 

  25. Park, J.H., Lee, T.H.: Synchronization of complex dynamical networks with discontinuous coupling signals. Nonlinear Dyn. 79, 1353–1362 (2015)

    Article  MathSciNet  Google Scholar 

  26. Mikheev, Y., Sobolev, V., Fridman, E.: Asymptotic analysis of digital control systems. Autom. Remote Control 49, 1175–1180 (1988)

    MathSciNet  MATH  Google Scholar 

  27. Astrom, K., Wittenmark, B.: Adaptive Control. Addison-Wesley, Reading (1989)

    Google Scholar 

  28. Liu, K., Fridman, E.: Wirtinger’s inequality and Lyapunov-based sampled-data stabilization. Automatica 48, 102–108 (2012)

    Article  MathSciNet  Google Scholar 

  29. Seuret, A., Briat, C.: Stability analysis of uncertain sampled-data systems with incremental delay using looped-functionals. Automatica 55, 274–278 (2015)

    Article  MathSciNet  Google Scholar 

  30. Lee, T.H., Park, J.H.: Stability analysis of sampled-data systems via free-matrix-based time-dependent discontinuous Lyapunov approach. IEEE Trans. Autom. Control 62, 3653–3657 (2017)

    Article  MathSciNet  Google Scholar 

  31. Lee, T.H., Park, J.H.: Improved criteria for sampled-data synchronization of chaotic Lur’e systems using two new approaches. Nonlinear Anal. Hybrid Syst. 24, 132–145 (2017)

    Article  MathSciNet  Google Scholar 

  32. Lee, T.H., Park, J.H.: New methods of fuzzy sampled-data control for stabilization of chaotic systems. IEEE Trans. Syst. Man Cybern. Syst. (2018).

    Article  Google Scholar 

  33. Zhang, H., Cheng, P., Shi, L., Chen, J.: Optimal denial-of-service attack scheduling with energy constraint. IEEE Trans. Autom. Control 60, 3023–30288 (2015)

    Article  MathSciNet  Google Scholar 

  34. Wang, Z., Wang, D., Shen, B., Alsaadi, F.E.: Centralized security-guaranteed filtering in multirate-sensor fusion under deception attacks. J. Franklin Inst. 355, 406–420 (2018)

    Article  MathSciNet  Google Scholar 

  35. Teixeira, A., Shames, I., Sandberg, H., Johansson, K.H.: A secure control framework for resource-limited adversaries. Automatica 51, 135–148 (2015)

    Article  MathSciNet  Google Scholar 

  36. Sandberg, H., Amin, S., Johansson, K.H.: Cyberphysical security in networked control systems: an introduction to the issue. IEEE Control Syst. Mag. 35, 20–23 (2015)

    MathSciNet  Google Scholar 

  37. Zhu, M., Martinez, S.: Stackelberg game analysis of correlated attacks in cyber-physical system. In: Paper Presented at the American Control Conference, 29 June–1 July San Francisco, pp. 4063–4068 (2011)

    Google Scholar 

  38. Yuan, Y., Zhang, P., Guo, L., Yang, H.J.: Towards quantifying the impact of randomly occurred attacks on a class of networked control systems. J. Franklin Inst. 354, 4966–4988 (2017)

    Article  MathSciNet  Google Scholar 

  39. Ding, D., Han, Q.-L., Xiang, Y., Ge, X., Zhang, X.: A survey on security control and attack detection for industrial cyber-physical systems. Neurocomputing 275, 1674–1683 (2018)

    Article  Google Scholar 

  40. Sun, Y.-C., Yang, G.-H.: Periodic event-triggered resilient control for cyber-physical systems under Denial-of-Service attacks. J. Franklin Inst. (2018).

    Article  MathSciNet  MATH  Google Scholar 

  41. Liu, J., Gu, Y., Cao, J., Fei, S.: Distributed event-triggered \(\mathcal{H}_{\infty }\) filtering over sensor networks with sensor saturations and cyber-attacks. ISA Trans. (2018).

    Article  Google Scholar 

  42. Zhang, W., Wang, Z., Liu, Y., Ding, D., Alsaadi, F.E.: Sampled-data consensus of nonlinear multiagent systems subject to cyber attacks. Int. J. Robust Nonlinear Control 28, 53–67 (2018)

    Article  MathSciNet  Google Scholar 

  43. Xu, W., Ho, D.W.C., Zhong, J., Chen, B.: Event/self-triggered control for leader-following consensus over unreliable network with DoS attacks. IEEE Trans. Neural Netw. Learn. Syst.

  44. Seuret, A., Gouaisbaut, F.: Wirtinger-based integral inequality: application to time-delay systems. Automatica 49, 2860–2866 (2013)

    Article  MathSciNet  Google Scholar 

  45. Park, P.G., Ko, J.W., Jeong, C.: Reciprocally convex approach to stability of systems with time-varying delays. Automatica 47, 235–238 (2011)

    Article  MathSciNet  Google Scholar 

  46. Gu, K., Kharitonov, V.-L., Chen, J.: Stability of Time-Delay Systems. Birkhauser Boston, Cambridge (2003)

    Book  Google Scholar 

Download references


The authors would like to thank the anonymous reviewers and the editor for their valuable comments and suggestions that helped improve the manuscript.


This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (Grant number NRF-2018R1C1B5036886) and also supported by research funds for newly appointed professors of Chonbuk National University in 2017.

Author information

Authors and Affiliations



All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Ju H. Park.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lee, T.H., Park, J.H. Design of sampled-data controllers for the synchronization of complex dynamical networks under controller attacks. Adv Differ Equ 2019, 184 (2019).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: