Anomaly detection using deep learning approach for IoT smart city applications

With the advancements of IoT devices, many smart applications start to rule this era. In particular, smart cities has been adapted and realized by many countries around the world. In smart cities, vas amount of data is generated at every second. This vast data need a transmission medium which could be wireless standard. However, security is the main concern in such applications since the smart transmission always binds with anomalies. The existing anomaly detection systems need improvement in accuracy due to inefficient feature extraction and selection procedure. This paper proposes an accurate anomaly detection technique that built upon deep learning approach. We proposed a Combined Deep Q-Learning (CDQL) algorithm for anomaly detection. Priory, optimal features are selected by using Spider Monkey Optimizer (SMO). With the optimal features, CDQL detects anomalies accurately. In addition, the CDQL algorithm learns the environment in order to monitor the network data continuously. This continuous monitoring and optimum features helps in accuracy improvement up to 98%.