Nowadays, the security of small-medium enterprises against cyber-attacks is a matter of great importance and challenging area, as they are financially and functionally affected. Novel and sophisticated attacks are emerging daily, targeting and threatening a large number of businesses in the world. For this reason, the implementation and optimization of the performance of Intrusion Detection Systems have attracted the interest of the scientific community. In this paper, a machine learning solution based on a deep neural network is proposed, in order to detect malicious behavior (DDoS and Malware cyber-threats) in the network traffic of an SME in real-time. The experimental results for the intrusion detection system showed that the proposed model can achieve high accuracy, having at the same time low false positive rate, while distinguishng between malicious and normal network traffic.

In modern communication networks, the integrity of the security is of great importance, since the existence of cyber attacks may lead to disastrous financial and social consequences. The anomaly detection constitutes an essential part of network security. This paper proposes a two-stage procedure to provide a solution regarding the anomaly detection and threat identification. The proposed method is suitable for modern communication networks and upcoming smart networks. The first stage of the method concerns the detection of abnormal incidents and the second stage involves the identification of the type of cyber threats, in case of an attack. The method based on the development of artificial neural network models and the UNSW-NB15 dataset is used to validate the proposed methodology. The experimental results confirm that the proposed method identifies all type of threats in comparison to the already known methods that identify only the threats that appear frequently.

No abstract