A novel network security situation assessment technology for zero-day attacks based on improved adversarial autoencoder
Runjie Liu, Yiyang Liu, Le Chen · 2023
In response to the difficulty in detecting and evaluating Zero-day Attack in the field of network security, this paper proposes a novel network security situation assessment technology based on deep learning. This research introduces a two-phase assessment model to achieve the detection and assessment of unknown attacks. The first phase focuses on reconstruction-based network situation anomaly detection, which is utilized to detect and assess anomalous traffic, including unknown attacks. In the second phase, a network attack identification system is developed to identify various types of known attacks. The overall security situation value is quantified by applying a weighted average to the results obtained from both phases. The method was validated on the public benchmark dataset UNSW-NB15, and the experimental results showed that the proposed technique has the ability to evaluate unknown Zero-day attacks, and the evaluation of known attacks is better than the baseline and existing models. By leveraging this technology, network security managers can gain a comprehensive understanding of the current threat landscape faced by the network. This empowers them to actively defend the network security system, mitigate the risk of unknown network attacks to system resources, and ensure the overall security of the network system.