Distributed denial-of-service attacks have become the major threat to Internet today. IP traceback is one of the most effective techniques to defeat these attacks by identifying attack sources even in the presence of IP spoofing. Because of low marking packet utilization, the convergence time of traditional probabilistic packet marking (PPM) schemes is still too long. In order to shorten the convergence time, a distributedlog- based IP traceback scheme is proposed. Theoretical analysis and simulation results show that this scheme not only can converge more quickly than previous PPM schemes, but also has much less log overhead than other log-based schemes.