1: \begin{abstract}

2:  This work studies the distributed resilient output containment control of heterogeneous multi-agent systems against composite attacks, including Denial-of-Services (DoS) attacks, false-data injection (FDI) attacks, camouflage attacks (CAs), and actuation attacks (AAs). Inspired by the concept of Digital Twin, a twin layer (TL) with  higher security and privacy is used to decouple the above problem into the defense protocol against DoS attacks on the TL and  the defense protocol against AAs on the cyber-physical layer (CPL). First, considering the existence of TL layer modeling error, we provide distributed observers to estimate the leaders dynamics for each follower. Then, distributed observers, which use the estimates of leaders dynamics by the above observers, are provided to observe the trajectories of followers  under DoS attacks. On the CPL, we give the estimated solutions to the output regular equations by estimating the  leaders dynamics. Moreover,

3:  decentralized adaptive attack-resilient control schemes are provided to deal with the unbounded AAs. Then, the output containment error can achieve uniformly ultimately bounded (UUB) convergence by the above control protocols and the boundary of output containment error can be given explicitly. Finally, the simulations show the effectiveness of these proposed control protocols.

4: \end{abstract}

5: