High availability in network services is crucial for effective largescale distributed computing. While distributed denial-of-service (DDoS) attacks through massive packet flooding have baffled researchers for years, a new type of even more detrimental attack—shrew attacks (periodic intensive packet bursts with low average rate)—has recently been identified. Shrew attacks can significantly degrade well-behaved TCP sessions, repel potential new connections, and are very difficult to detect, not to mention defend against, due to its low average rate. We propose a new stateful adaptive queue management technique called HAWK (Halting Anomaly with Weighted choKing) which works by judiciously identifying malicious shrew packet flows using a small flow table and dropping such packets decisively to halt the attack such that well-behaved TCP sessions can re-gain their bandwidth shares. Our NS-2 based extensive performance results indicate that HAWK is highly agile.