Wireless ad hoc networks realize end-to-end communications in a cooperative manner. In this paradigm, multiple nodes coordinate to form a multi-hop route, when communication needs to take place between a source and a destination that are not within communication range. Thus, intermediate nodes are willing to carry traffic other than their own. For ad hoc networks deployed in hostile environments, a protocol-compliant behavior on behalf of all nodes of the network cannot be assumed. Selfish and/or malicious users may misconfigure their devices to refuse forwarding any traffic, in order to conserve energy resources or degrade the network performance. We address the problem of identifying and isolating misbehaving nodes that refuse to forward packets. We develop a comprehensive system called Audit-based Misbehavior Detection (AMD) that effectively and efficiently isolates both continuous and selective packet droppers. The AMD system integrates reputation management, trustworthy route discovery, and identification of misbehaving nodes based on behavioral audits. It consists of three modules: the reputation module, the route discovery module, and the audit module. All three modules are tightly integrated to ensure that multi-hop communications take place over paths free from malicious nodes.