Wireless ad hoc networks rely on multi-hop routes to transport data from source to destination. The routing function is implemented in a collaborative manner, with each node responsible for relaying traffic to the destination. However, an increasingly sophisticated pool of users with easy access to commercial wireless devices, combined with the poor physical and software security of the devices, can lead to node misconfiguration or misbehavior. A misbehaving node may refuse to forward packets in order to conserve its energy (selfishness), or simply degrade network performance (maliciousness). In this paper, we investigate the problem of uniquely identifying the set of misbehaving nodes who refuse to forward packets. We propose a novel misbehavior identification scheme called REAct that provides resource-efficient accountability for node misbehavior. REAct identifies misbehaving nodes based on a series of random audits triggered upon a performance drop. We show that a source-destination pair using REAct can identify any number of independently misbehaving nodes based on behavioral proofs provided by nodes. Proofs are constructed using Bloom filters which are storage-efficient membership structures, thus significantly reducing the communication overhead for misbehavior detection.