MSc Dissertation - repository placeholder
Keywords: Intrusion Detection System, Machine learning algorithms, Deep learning algorithms, Deep Neural Network, clustering, supervised and unsupervised learning, CSE-CIC-IDS2018 dataset
Network related services, programs and applications are developing greatly, however, network security breaches are also developing with them. Network security is an evolving, challenging and a critical task. It is essential that there is a system in place to identify any harmful movement happening in network. An Intrusion detection system (IDS) has become the prerequisite software addressing cyber security in the modern era. Especially, with the greater complexity of advanced cyber-attacks and as such the uncertainty surrounding the detection of the types of attacks. This thesis proposes a novel approach using an ensemble of K-Means and Gaussian Mixture clustering combined with a deep neural network (DNN) algorithm. When compared with traditional artificial neural network’s (ANN’s) used within an IDS, our approach implements modern advances in deep learning such as initialising the parameters through the unsupervised pre-training clustering ensemble, therefore improving the detection accuracy. We hope our results will show that the proposed approach can provide a real-time response to the attack with a greatly increased detection ratio for false flags.
MSc Dissertation - repository placeholder
Keywords: Intrusion Detection System, Machine learning algorithms, Deep learning algorithms, Deep Neural Network, clustering, supervised and unsupervised learning, CSE-CIC-IDS2018 dataset
Network related services, programs and applications are developing greatly, however, network security breaches are also developing with them. Network security is an evolving, challenging and a critical task. It is essential that there is a system in place to identify any harmful movement happening in network. An Intrusion detection system (IDS) has become the prerequisite software addressing cyber security in the modern era. Especially, with the greater complexity of advanced cyber-attacks and as such the uncertainty surrounding the detection of the types of attacks. This thesis proposes a novel approach using an ensemble of K-Means and Gaussian Mixture clustering combined with a deep neural network (DNN) algorithm. When compared with traditional artificial neural network’s (ANN’s) used within an IDS, our approach implements modern advances in deep learning such as initialising the parameters through the unsupervised pre-training clustering ensemble, therefore improving the detection accuracy. We hope our results will show that the proposed approach can provide a real-time response to the attack with a greatly increased detection ratio for false flags.
MIT License