Volume 16, No 2, 2019
DSVM-FCNN: Using The DL Approach To Develop Intrusion Detection Systems
Mashail Saleh Ibrahim Alsalamah
Abstract
The number of people using online and networked resources has steadily risen over the past few years. Daily, a significant amount of data is produced over the Internet at speeds ranging from zeta to petabytes. Due to their size, diversity, velocity, and veracity, these may be classified as big data. Network, Internet, website, and enterprise security concerns are expanding as quickly as their usage. Even though several AI/ML-based intrusion-detection systems (IDSs) have been created for various network assaults, the great majority of these systems cannot either identify unknown assaults or react to them in a timely manner. Deep learning (DL) algorithms are increasingly used for extensive big data analysis. However, although they generally demonstrate excellent performance, their capacity to detect intrusions in big data environments has not been explored. The purpose of this study is to suggest an effective IDS based on distributed support vector machine with fully connected neural networks (DSVM-FCNN) using a DL approach that is aware of big data. Concerning incoming traffic packets, this model can identify intricate relationships and long-term interdependence. This would allow us to improve the proposed intrusion detection system's accuracy while decreasing the number of false alerts. In addition, the DL algorithms presented in this research, which suffer from slow execution speed as a result of their enormous complexity, may be made faster by the use of big data analytical approaches. Experimental results show that the proposed method is far better when compared to conventional IDS systems.
Pages: 363-380
Keywords: Big data, deep learning, DSVM-FCNN, Network intrusion detection systems.