Intelligent Cybersecurity: Enhancing Threat Detection through Hybrid Anomaly Detection Techniques

Abstract

Conventional methods prove to be inefficient in identifying new and developed threats, so it is crucial to find ways to improve the situation. This paper presents an extended anomaly detection model that combines features of both machine learning and statistical methods to improve threat detection's accuracy and versatility. The advantage of such powerful methods of hybrid detection is that it is planned to increase the probability of detection of not only known threats but also such threats that may be unfamiliar to the anti-virus program. A simulation was carried out to validate the model; then, in an actual 'live' working scenario, the model proved efficient, quickly identifying threats with fewer false alarms. Conceptual tools of different types help represent the accuracy and the response rate of the system. Furthermore, in this work, problems like high computational cost and quality of the input data are investigated, with solutions provided for improvement. The results demonstrate that both single-discipline and mixed anomaly detection approaches can provide a valuable contribution towards the enhancement of the cybersecurity situation in various types of networks.