The fifth-generation (5G) network is presented as one of the main options for Industry 4.0 connectivity. Ultra-Reliable and Low Latency Communications (URLLC) is the 5G service category used by critical mechanisms, with a millisecond end-to-end delay and reduced probability of failure. 5G defines new numerologies, together with mini-slots for a faster scheduling. The main challenge of this is to select the appropriate numerology according to radio conditions. This fact is very important in industrial scenarios, where the fundamental problems are interference and multipath propagation, due to the presence of concrete walls and large metallic machinery and structures. Therefore, this paper is focused on analyzing the impact of the numerology selection on the delay experienced at radio link level for a remote-control service. The study, which has been carried out in a simulated cellular factory environment, has been performed for different packet sizes and channel conditions, focusing on outliers. Evaluation results show that not always a higher numerology, with a shorter slot duration, is appropriate for this type of service, particularly under Non-Line-of-Sight (NLOS) conditions.

Industry 4.0 is being adopted by the manufacturing sector to improve the flexibility and reduce installation costs by the use of wireless connectivity. There is an open question of which wireless technology deployment should be used in the factory to fulfil the requirements for next-generation applications such as autonomous mobile robots. Wi-Fi technology is the most extended and easy to deploy, while the fifth generation of mobile networks (5G) network has been designed to support these industrial needs. Therefore, it is important to compare both technologies from a performance point of view, especially under different load conditions and number of devices. The use of multi-connectivity between 5G and Wi-Fi can also be an option to fulfil the requirements for the most critical real-time applications. In this paper, we empirically measure the scalability of 5G, Wi-Fi and multi-connectivity in the “Aalborg University 5G Smart Production Lab” and compare them in terms of latency and packet loss with different packet sizes. We found that in general Wi-Fi obtains lower latencies but large tails in the distribution, with a higher packet loss compared to 5G. On the other hand, 5G latency is very consistent with bounded tails, and low packet loss is obtained. In terms of scalability, 5G scales better than Wi-Fi, the latter being very affected by the number of devices transmitting data. Finally, multi-connectivity showed an improved reliability and lower latencies in all evaluated cases.

The fifth-generation (5G) network is presented as one of the main options for Industry 4.0 connectivity. To comply with critical messages, 5G offers the Ultra-Reliable and Low latency Communications (URLLC) service category with a millisecond end-to-end delay and reduced probability of failure. There are several approaches to achieve these requirements; however, these come at a cost in terms of redundancy, particularly the solutions based on multi-connectivity, such as Packet Duplication (PD). Specifically, this paper proposes a Machine Learning (ML) method to predict whether PD is required at a specific data transmission to successfully send a URLLC message. This paper is focused on reducing the resource usage with respect to pure static PD. The concept was evaluated on a 5G simulator, comparing between single connection, static PD and PD with the proposed prediction model. The evaluation results show that the prediction model reduced the number of packets sent with PD by 81% while maintaining the same level of latency as a static PD technique, which derives from a more efficient usage of the network resources.

The fifth-generation technology is called to support the next generation of wireless services and realize the “Internet of Everything” through Machine-Type Communications and Cellular Internet of Things optimizations. As part of these optimizations, Release 15 introduced a new mechanism, known as Early Data Transmission (EDT), that allows the transmission of data during the Random Access procedure. This feature, intended particularly for infrequent and small data transmissions, aims to reduce the latency and the power consumption of user equipments. Nonetheless, despite the importance of this novelty and the general agreement about its effectiveness, there are few papers in the literature that provide insight into its implementation and analyze the advantages and disadvantages of its two different implementation options (Control and User Plane). Moreover, although security is recognized as a crucial aspect for the correct deployment of this technology, we have not found any paper that reviews the security issues and features of this mechanism. As a consequence of such a lack of papers and the complexity of mobile network protocols, there is a divide between security experts and EDT researchers, that prevents the easy development of security schemes. To overcome this important gap, this paper offers a tutorial of EDT and its security, analyzing its main vulnerabilities and concluding with a set of recommendations for researchers and manufacturers. In particular, due to the simplifications in the protocols done by EDT, vulnerabilities such as packet injection, replay attacks and injection of fake values to disable EDT have been found.

The ever growing demand for services with high requirements is a significant force behind the evolution of cellular networks. End-to-End (E2E) analysis is a key methodology for the assesment of service performance, often paired with Machine Learning (ML) algorithms. However, ML comes with its own security challenges. This work will study attaks on the data, some of the major security challenges in ML systems. Namely, it will study poisoning and evasion in ML regressors, offering different perspectives on their impact in the outcome of the system within a mobile network and the effects on user perception. Moreover, new metrics for computing the statistical variation of model predictions are proposed to evaluate the attack effects. A realistic testbed is used to evaluate the proposed mechanisms.