In this dissertation, we describe the portable, open-source distributed simulation framework (WINDS) targeting simulations of wireless network infrastructures that we have developed. We present the simulation framework which uses modular architecture and apply the framework to studies of mobility pattern effects, routing and intrusion detection mechanisms in simulations of large-scale wireless ad hoc, infrastructure, and totally mobile networks. The distributed simulations within the framework execute seamlessly and transparently to the user on a symmetric multiprocessor cluster computer or a network of computers with no modifications to the code or user objects. A visual graphical interface precisely depicts simulation object states and interactions throughout the simulation execution, giving the user full control over the simulation in real time. The network configuration is detected by the framework, and communication latency is taken into consideration when dynamically adjusting the simulation clock, allowing the simulation to run on a heterogeneous computing system. The simulation framework is easily extensible to multi-cluster systems and computing grids. An entire simulation system can be constructed in a short time, utilizing user-created and supplied simulation components, including mobile nodes, base stations, routing algorithms, traffic patterns and other objects. These objects are automatically compiled and loaded by the simulation system, and are available for dynamic simulation injection at runtime. Using our distributed simulation framework, we have studied modern intrusion detection systems (IDS) and assessed applicability of existing intrusion detection techniques to wireless networks. We have developed a mobile agent-based IDS targeting mobile wireless networks, and introduced load-balancing optimizations aimed at limited-resource systems to improve intrusion detection performance. Packet-based monitoring agents of our IDS employ a CASE-based reasoner engine that performs fast lookups of network packets in the existing SNORT-based intrusion rule-set. Experiments were performed using the intrusion data from MIT Lincoln Laboratories studies, and executed on a cluster computer utilizing our distributed simulation system.