Intrusion Detection and Prevention

Nowadays, the global internet consists of billions of devices, the number of which is constantly growing. This trend is due to the increasing use of consumer electronics, where more and more sensors are installed every day. These devices have limited computing resources, and in most cases, information management is transferred to external devices. The latest devices, in turn, connect with each other and create communication channels to transmit time and status data. Existing 4G mobile networks cannot provide the necessary capabilities for the continued development of the networks of the Internet of Things. Therefore, we can immediately conclude that 5G networks should become the basis of the next generation of large-bandwidth data transmission networks.

Fifth-generation cellular networks forced the implementation of 5G and beyond networks, which offer capacity expansion strategies to handle great connectivity issues and can offer very high throughput and low-latency. 5G and beyond technology uses IoT, AI/ML, and blockchain, and its goal is to establish secure and reliable UAV networks. Therefore, the big work must be conducted to ensure security of 5G and beyond networks. It can be very relevant to the integration of protected mechanisms, which use machine learning (ML) and artificial intelligence (AI) techniques. Scientists apply ML algorithms in the development of IDS systems in order to identify and classify malicious traffic. It is also important to identify the possible threats of 5G and beyond networks in real-time. The main problem with the research is that the security system must work very quickly, the processing delay should be very small, and in other cases, it is not efficient to use the system for real-time threats, identification, and classification.

The intrusion detection system, which is presented in this paper, addresses the stringent design, implementation, and deployment aspects of high-bandwidth 5G network cores. Thus, traditionally, the data traffic is filtered mostly using semiautomatic approaches. These usually generate low levels of data pattern detection accuracy, and they are not able to adapt and detect unknown data patterns. It is important to mention that the integrated intrusion detection system, which is presented in this paper, is one of the very few relevant systems that are proven to detect known and unknown threat patterns in a large 5G network core with high accuracy and without interfering with the low-latency levels of the implied data network, as they are perceived by the end users.

It is relevant to note that 5G networks create broad-bandwidth channels. However, the increased efficiency that these new networks create is largely due to the number of intelligent devices supported and the related applications. Broad-bandwidth data links: intelligent application deployments require data links capable of a minimum of 25 Mbps and are designed to sustain meaningful augmented reality (AR) and virtual reality (VR) and data containers. Large scale and structurally flexible networks: This is defined by the network function virtualization (NFV) mechanism to create the networks needed. 5G IoT low-latency data networks are designed to sustain intelligent applications that need to transmit and receive data in real-time and use communication channels with delays of no more than five milliseconds. Safety and fault tolerance: due to the presence of significantly fewer base stations in the 5G network, handoffs must be done when maintaining optimal coverage of the network. Data privacy and protection: applications that work with sensitive data, such as patient personal information, need mechanisms to prevent any unauthorized access attempts. Battery life: mobility is central to 5G data networks; hence, energy efficiency must be taken into account. Connectivity: 5G data networks must offer simultaneous, stable access to a huge number of devices, which means making the right design and implementation decisions. Mobility: This requirement supplements the need to create the right environment for the development of many devices that require reliable mobile data links. It must be noted that although smart devices that need to work on 5G networks must handle huge amounts of data, they do not, in most cases, have sufficient hardware resources to process the given data. Therefore, in most cases, information processing is transferred to systems in the cloud that extract useful information from unprocessed data by considering data analysis techniques.

The scientists discussed the security problems of 5G networks. The following problems are identified: 1. As the 5G data network's architecture mainly uses software configurations, it has a much bigger exposure to attacks on the software. The attacks can be designed using existing security flaws or bugs. These attacks can influence the operation of 5G data networks. 2. The architecture of 5G data networks includes novel models and functions, because of this, they can be the target of hackers' attacks. The functions of key management for the networks and the base station can become the target of attacks. 3. The majority of 5G operators of mobile networks depend on suppliers; this fact can lead to additional attacks on 5G data networks. The impact of such attacks will also be greatly increased. 4. 5G data networks will include a huge amount of different smart devices. Therefore, attacks such as DOS and DDOS will become much more relevant and often. 5. The key feature of 5G network slicing can also be considered a security problem. The attackers can force the service to use the slice that was not intended for it.

It must be mentioned that the vulnerabilities of injecting malicious code into the system of 5G networks were also identified.

The contribution that is presented in this paper relates to the following perspectives.

• This article presents the current advances in the improvement of standards that simulate 5G networks through virtualized infrastructures. This determines a significant improvement in the telecommunications operators' administrative and infrastructure maintenance costs.

• This paper presents the experience that the authors gained when deploying the described system on the core infrastructure of a major telecommunications operator.

• This paper outlines the relevant research areas and challenges that are likely to affect the efficient design and implementation of large 5G data networks, which use secure and economically efficient virtualized infrastructures.

• The performance evaluation of the system, which considered a comprehensive sample of real networking data, demonstrates that the system is capable of detecting unknown and existing malicious data traffic patterns in a timely manner with a high level of accuracy.

• To the best of our knowledge, this is one of the very few machine-learning-based intrusion detection systems that is compatible with the proper and timely detection of malicious data traffic patterns in large broadband 5G data networks.

The rest of this paper is organized according to the following structure.: First, the essential materials and methods are described. Following, the architecture of the intrusion detection system is described, and the fundamental algorithmic and implementational features are presented. Following this, the real-world performance of the implemented system is thoroughly assessed through a comprehensive case study. The last section presents the planned development directions. It also concludes the paper.