In this dissertation, several novel cyber fault diagnosis and prognosis and defense methodologies for cyber-physical systems have been proposed. First, a novel routing scheme for wireless mesh network is proposed. An effective capacity estimation for P2P and E2E path is designed to guarantee the vital transmission safety. This scheme can ensure a high quality of service (QoS) under imperfect network condition, even cyber attacks. Then, the imperfection, uncertainties, and dynamics in the cyberspace are considered both in system model and controller design. A PDF identifier is proposed to capture the time-varying delays and its distribution. With the modification of traditional stochastic optimal control using PDF of delays, the assumption of full knowledge of network imperfection in priori is relaxed. This proposed controller is considered a novel resilience control strategy for cyber fault diagnosis and prognosis. After that, we turn to the development of a general framework for cyber fault diagnosis and prognosis schemes for CPSs wherein the cyberspace performance affect the physical system and vice versa. A novel cyber fault diagnosis scheme is proposed. It is capable of detecting cyber fault by monitoring the probability of delays. Also, the isolation of cyber and physical system fault is achieved with cooperating with the traditional observer based physical system fault detection. Next, a novel cyber fault prognosis scheme, which can detect and estimate cyber fault and its negative effects on system performance ahead of time, is proposed. Moreover, soft and hard cyber faults are isolated depending on whether potential threats on system stability is predicted. Finally, one-class SVM is employed to classify healthy and erroneous delays. Then, another cyber fault prognosis based on OCSVM is proposed.