Attacks against broadcast services in wireless sensor networks will have the most catastrophic effects on power and other resource consumptions. As bogus broadcast packets are propagated through sensor nodes without being filtered out, sensor nodes will waste their resources such as power and memory on transmitting and buffering those false packets. In networks where complex Denial-of-Service (DoS) attacks may be launched, each node should be able to limit its effects on resource consumptions. In addition, it is strongly desired to identify compromised sensor node(s) which has injected invalid packets (intrusion detection) and to isolate the nodes (remediation) such that no more DoS attack is possible through the compromised sensor node(s) in the future. A new secure broadcast authentication scheme for sensor networks is proposed in this paper based upon predictive hashing and μTESLA techniques. This new scheme may be used to detect invalid packets as quickly as possible and to isolate links that are from the compromised nodes, which results in enhanced resistance to various security attacks including DoS attacks. We present the details of the scheme along with the simulation results.