Security threats in water distribution systems have predominantly increased in recent times. In addition to the traditional security measures like data authentication and encryption employed in these systems, there is a need to adopt anomaly detection techniques to confront the interruptions caused due to various attacks. It is paramount that the anomaly detection techniques be used not only detect the attacks with more accuracy but also be capable enough to perform real-time attack detection. Machine learning and deep learning methods are widely employed as attack detection methods. The anomaly detection technique proposed in this work is based on deep neural network architectures, such as convolutional neural networks (CNN) and recurrent neural networks (RNN). Different variants of CNN and RNN techniques are applied to the BATtle of the Attack Detection ALgorithms (BATADAL) dataset and the performance comparison based on the obtained results demonstrated that the multivariate CNN-long-short term memory network model outperformed the other variants of CNN and RNN methods.