| Summary: | In recent years, the explosive growth in internet and technology use has led to an alarming escalation in both the frequency and severity of cyberattacks. As such, proactive detection and prevention of these attacks have become a matter of paramount importance. This need for vigilance has catalyzed the adoption of Machine Learning (ML) and Deep Learning (DL) techniques to effectively identify and analyze network traffic content, predict potential cyberattacks, and respond promptly to these security threats. ML and DL methods offer innovative solutions by facilitating the development of sophisticated models that meticulously analyze patterns in network traffic data. By identifying deviations from expected behaviors, these techniques enable the early detection and prevention of impending attacks. They achieve this by learning from the data, improving their ability to detect attacks over time, and responding effectively to new, unseen threats. However, contemporary intrusion detection methods face significant challenges, particularly related to imbalanced classes, low detection rates, and high false alarm rates. Addressing these hurdles is critical for the development of a robust and efficient intrusion detection system. Our research seeks to confront these issues head-on, by proposing an innovative, high-performance intrusion detection system tailored specifically to handle imbalanced datasets. Our methodology not only offers improvements in detection rates and false alarm rates but also provides a feasible solution for handling class imbalance in the data. We anticipate that our findings will pave the way for more robust cyber defense mechanisms in this era of ever-evolving security threats. © 2023 IEEE.
|
|---|