This paper proposed an ARM-CPD scheme that is a simple but fast and effective approach to detect SYN flooding attacks. Instead of managing all real time ongoing traffic on the network, ARM-CPD only monitors the SYN packet and use it to predict the SYN packet in the near future to detect the SYN flooding attacks. To get the prediction SYN traffic, the Autoregressive Integrated Moving Average Model (ARIMA) is proposed; and to make the detection method insensitive to site and access pattern, a non-parametric Cumulative Sum (CUSUM) algorithm is applied. The trace-driven simulations demonstrate that ARM-CPD can shorten the detection time of SYN flooding attack effectively.