In the fast-progressing field of wireless communications, the forthcoming 6G networks are expected to revolutionize the way we communicate, offering unparalleled speed, minimal latency, and seamless connectivity. However, amid this evolution, the paramount concern remains the security and privacy of the data traversing these networks. Traditional centralized artificial intelligence (AI) techniques already struggle to keep up with the vast amount of data of future 6G networks and deal with the increasing worries about privacy. Federated learning (FL), emerges as a key enabler of Trustworthy AI (TAI), empowering the engagement of distributed network nodes in AI training without the need for exchanging raw data, thereby mitigating the risks associated with centralized data processing. In this paper, we provide a comprehensive survey on the potential of FL in enhancing the security of 6G networks. Particularly, we begin by providing the necessary background on 5G networks and FL, setting the stage for understanding their current and future implications. We then explore the current state-of-the-art of FL applications within 5G networks and their relevance to the future threat landscape of 6G. Subsequently, we examine the inherent vulnerabilities of FL systems, major attacks against FL in the context of 5G networks, and corresponding defense mechanisms. Finally, we discuss the integration of advanced FL technologies and concepts towards enhanced cybersecurity and privacy in 6G networks, aiming to cover all aspects and future perspectives of FL within the context of the forthcoming 6G threat landscape.

The rapid advancement of Artificial Intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL), has produced high-performance models widely used in various applications, ranging from image recognition and chatbots to autonomous driving and smart grid systems. However, security threats arise from the vulnerabilities of ML models to adversarial attacks and data poisoning, posing risks such as system malfunctions and decision errors. Meanwhile, data privacy concerns arise, especially with personal data being used in model training, which can lead to data breaches. This paper surveys the Adversarial Machine Learning (AML) landscape in modern AI systems, while focusing on the dual aspects of robustness and privacy. Initially, we explore adversarial attacks and defenses using comprehensive taxonomies. Subsequently, we investigate robustness benchmarks alongside open-source AML technologies and software tools that ML system stakeholders can use to develop robust AI systems. Lastly, we delve into the landscape of AML in four industry fields –automotive, digital healthcare, electrical power and energy systems (EPES), and Large Language Model (LLM)-based Natural Language Processing (NLP) systems– analyzing attacks, defenses, and evaluation concepts, thereby offering a holistic view of the modern AI-reliant industry and promoting enhanced ML robustness and privacy preservation in the future.

Short-term load forecasting (STLF) is crucial for the daily operation of power grids. However, the non-linearity, non-stationarity, and randomness characterizing electricity demand time series renders STLF a challenging task. Various forecasting approaches have been proposed for improving STLF, including neural network (NN) models which are trained using data from multiple electricity demand series that may not necessarily include the target series. In the present study, we investigate the performance of a special case of STLF, namely transfer learning (TL), by considering a set of 27 time series that represent the national day-ahead electricity demand of indicative European countries. We employ a popular and easy-to-implement feed-forward NN model and perform a clustering analysis to identify similar patterns among the load series and enhance TL. In this context, two different TL approaches, with and without the clustering step, are compiled and compared against each other as well as a typical NN training setup. Our results demonstrate that TL can outperform the conventional approach, especially when clustering techniques are considered.