Executive Summary : | The project investigates two collective behaviours of networked dynamical systems, also known as multi-agent systems. The behaviours under investigation are called formation and containment control, which have many civilian and military applications. However, to realize these control tasks, instead of a centralized controller monitoring and operating all the agents, each agent exchanges its state information with other communicating agents over a communication network to update their state values, which forms the basis of distributed control. Nonetheless, the agents sharing their information, which are very sensitive in these applications, over a network introduces vulnerability from cyber attacks like monitoring the data, or manipulating it, or may even disrupting its flow, thereby precluding from realizing the behaviours. In some cases, in the large-scale MAS, a few agents might be adversaries themselves, sending spurious signals to communicating agents to disturb the behaviours. To that end, this proposal aims at designing certain preventive secure control algorithms for these collective behaviours like 'differential private' and 'encrypted control' schemes which provide protection against monitoring/disclosure attack like eavesdropping, with an objective to assess the behaviour performance versus the security level achieved with these designs. Also, in reality, various types of attacks like disruption attack (e.g., denial-of-service) impeding the information flow and deception attacks (e.g. false data injection) altering the information can be launched in parallel on MAS network, where the timings of these attacks are independent to each-other. In this regard, through this proposal, the objective will be to design resilient control schemes that can address and minimize the effects of these hybrid attacks launched in parallel, but asynchronously. |