Multi Core DN​N based IDS for Botnet Attacks using KPCA Reduction Techniques

Abstract

Research on network security has recently acquired attention in the field of the Internet of Things. In the context of security, most of the IoT devices with the internet are connected directly which results in the exploitation of private data. Nowadays, the fraudster will release novel attacks very frequently especially for IoT devices. As a result, the traditional sophisticated Intrusion Detection System (IDS) model is not suitable for the identification of vulnerabilities in IoT devices. In our research work, we propose MCDNN for IDS. MCDNN is Multi Core DNN with having parallel optimizer. Rather than a traditional dataset, this paper experiment is conducted on the BoTIoT dataset. Since IoT devices generate a huge volume of data, this work focuses on reducing huge datasets using Kernel Principal Component Analysis(KPCA) reduction technique with optimizer parallelly. To decrease false alarm rate and maintaining less computational power multi-core is introduced in our research work. This helps identification of vulnerabilities in IoT devices using deep learning techniques faster. Experimental results indicate that designing MCDNN based IDS with different optimizers parallelly achieved higher performance than those of other techniques.