ABSTRACT
Many enterprises have started using cloud systems and services to enhance their productivity by migrating their applications, data, and infrastructures to cloud platforms. However, these kinds of migrations raise the number of attacks on cloud systems. Protecting these systems becomes an essential task for cloud providers based on intrusion detection system (IDS). Detecting normal and abnormal network traffic packets is the main task for any IDS. In this work, we introduce a hybrid intelligent IDS system based on long short-term memory (LSTM) networks and binary particle swarm optimization (BPSO). BPSO is used as a wrapper feature to reduce the high dimensionality of collected data and select the most valuable features. While LSTM is employed as a binary classifier. A real-time public dataset is used in this work called UNSW-NB15 to evaluate the proposed system. The obtained results show the proposed system detects abnormal network traffic with 92% accuracy.
Keywords: cloud service, network intrusion detection system, feature selection, swarm optimization, LSTM
