Detecting anomalies in unusual vessel movements is one of key activities to make a proper and timely decision in maritime logistics. Today, vessel trajectory data is provided through an automatic identification system (AIS) in near real time, but research into an algorithm for effectively identifying abnormal vessel movement through such big data is insufficient. This paper presents a new anomaly detection method, VAE-CUSUM, using AIS data through combining Variational Autoencoder(VAE) and CUSUM control chart. VAE-CUSUM provides an effective way of employing both VAE and control charts to monitor real-time movements of vessels, examining whether the vessel movements are being deviated from normality. The effectiveness of the proposed method is evaluated using both simulated data and real AIS data from maritime logistics.