Intrusion Detection and Prevention

Conclusions and Future Work

The 5G networks have support for real-world applications. These applications provide enough potential to form the foundation of an ever-connected community. However, there are still problems with their design, launch, and implementation, which interest all aspects of 5G network research. One of the most important of these research problems pertains to the timely detection of unauthorized access attempts, especially in the case of commercial networks. Thus, this article presents the current results of research carried out on this very important topic. This article also describes a real-time intrusion detection and prevention system based on machine learning. The system was tested by monitoring real-time 5G data traffic on the network of one of the leading Romanian telecommunications service providers using real data. The system was checked in terms of the concept of symmetric and asymmetric communication scenarios. The research shows that it is possible to develop a software system that blocks illegal traffic in real-time on a 5G data network. The tests on the real-world 5G network show us that the system can detect and classify newly arrived data/malicious data and make decisions in milliseconds, which reflects the computational efficiency and novelty of the offered product. In addition, the article describes, in an analytical form, the contribution that is related to the topic under discussion, which analyzes the existing problems and presents possible ways to solve them. The idea of this system can also be used for future generations of networks. The big advantage of this system is that it can work in real-time; however, the system also has limitations. It must be mentioned that the received accuracy score is 0.9414, which is rather low and must be increased. This is the big limitation of our system, and it must be improved. We suggest the use of data augmentation techniques to increase the accuracy score. We plan to improve the algorithm in order to increase the accuracy score. It could be interesting for other researchers to contribute to our research in order to increase the corresponding accuracy metric. In the future, we also suggest the addition of post-quantum encryption to the identification stage of 5G networks, which requires asymmetric encryption. We also suggest the addition of artificial intelligence to 5G and beyond networks in order to identify if the attack comes from a classical or quantum computer. Finally, we want to offer the novel model of 5G security, which will contain needed security mechanisms. The effective real-world relevance of the presented intrusion detection system is sufficiently justified by the requirement to automatically process 5G data traffic flows using self-developing data traffic analysis algorithmic cores and also by its demonstrated efficiency relative to a real-world deployment. Consequently, any further improvement of its detection accuracy and computational efficiency is implicitly useful.