AI at the network edge promises real-time perception and decision-making in safety-critical domains such as aerial robotics, autonomous vehicles, and 5G-enabled infrastructures. Yet, operating under resource constraints, dynamic, and adversarial conditions exposes edge AI systems to fragility, inefficiency, and security risks that threaten their safe operation. GuardAI, a Horizon Europe project, introduces a framework for resilient and trustworthy edge AI that unites three pillars: adversarial robustness, context-enhanced inference, and security-by-design. Initial project results include a diffusion-based adversarial purification framework optimized for real-time operation, lightweight deep unrolling architectures for LiDAR super-resolution with built-in outlier removal, and robust uncertainty quantification modules to improve confidence calibration. It further develops a context-enhanced inference engine that integrates visual, spatial, and operational context across multi-agent systems, and a risk-aware defense recommender that autonomously selects mitigation strategies based on evolving threat landscapes. Through representative Use Cases, covering monitoring with Unmanned Aerial Vehicle, decentralized 5G network analytics, and secure perception in connected autonomous vehicles, GuardAI demonstrates how robust and adaptive AI can be achieved within stringent edge constraints. Together, these technologies lay the groundwork for a new generation of secure, context-aware, and certifiable AI systems that can be trusted to operate autonomously in the physical world.

Network connectivity exposes network infrastructure and assets to vulnerabilities exploitable by attackers. Safeguarding these assets necessitates implementing security countermeasures. However, deploying countermeasures incurs various costs, including preparation and deployment time. Therefore, an Intrusion Response System (IRS) must consider both security and Quality of Service (QoS) costs when dynamically selecting countermeasures to address detected attacks. To address this challenge, we introduce a joint Security-vs-QoS optimization problem akin to the Weighted Set Cover Problem (WSCP), which is NP-complete. We propose two learning-based solutions leveraging Multi-Objective Reinforcement Learning and Deep Q-learning to navigate the security and QoS cost trade-off. Through extensive simulations under diverse settings, we validate the performance of our proposed solution, compare it with benchmark methods, and evaluate it using a project-derived 5G cybersecurity dataset.

This paper addresses the need for an explainable and generalized intrusion detection system (IDS) for the in-vehicle networks (IVNs). While machine learning (ML)-based IDS solutions show promising performance, there are still some challenges, such as the lack of trustworthiness and scarcity of attack representing data, hindering their adoption in the automotive cybersecurity. To address these issues, this paper proposes a centralized ML model training and decentralized execution-based framework, namely ACHILLES, that facilitates an explainable and generalizable automotive IDS. Under ACHILLES, different ML models can be trained centrally to enhance decentralized and onboard intrusion detection performance with multiple automotive datasets. In addition, we generate standard feature formats to assess the ML model’s generalization efficacy, where the quality of generalization and explainability is evaluated with SHapley Additive exPlanations (SHAP) by identifying the importance of the feature. We also propose a meta-learning scheme to construct suitable ML models trained by the proposed standard feature formats. The proposed feature format exhibits significant performance gain during ML model training and testing with four state-of-the-art controller area network (CAN)-bus datasets containing real, advanced attacks. The experimental results indicate that developing ML models using the generated generalized features and the meta learning-based model building process leads to enhanced performance. In particular, under the dataset cross train-test setting, the proposed feature format enhances the average accuracy by 40.1% for the baseline model, 32.4% for the meta-learned DNN, and 23.6% for the meta-learned Random Forest, compared with the baseline feature format.

Programmable Wireless Environments (PWEs) leverage Reconfigurable Intelligent Surfaces (RISes) to actively shape electromagnetic (EM) propagation, enabling advanced control over wireless channels. Beyond improved performance in B5G/6G networks, this control also introduces new security capabilities. Exploiting this, we propose RF-Fencing: a service that selectively suppresses EM signals toward eavesdroppers while preserving reliable communication for legitimate users, thereby significantly enhancing network covertness. Building on that, in this paper, we introduce SHIELD, the first RF-Fencing algorithm that partitions the PWE into Signal Suppression Areas (SSAs) and Signal Delivery Areas (SDAs) through on-the-fly merging of RIS configurations. Extensive EM analysis confirms SHIELD’s effectiveness in preventing wardens from intercepting critical information and achieving covert communications with minimal impact on legitimate users. Moreover, SHIELD can serve also as a jamming-mitigation mechanism and is applicable across various frequency bands and RIS designs.

Programmable Wireless Environments aim to render the communication environment a controllable, software-defined medium. Reconfigurable Intelligent Surfaces (RISes) are the key enabling technology, which can offer the real-time capability to manipulate impinging waves. RISes are expected to be widely deployed in B5G/6G networks to serve a large number of users simultaneously. Despite numerous analyses highlighting the benefits of utilizing previously unexploitable propagation factors through the use of RISes, there is a lack of analysis regarding their relation to the concept of network resource, their allocation to users/stakeholders and their fair pricing. Thus, this paper models RISes as networked resources. Based on this definition, the PRIME algorithm is proposed, the first algorithm for RIS resource allocation and joint pricing. PRIME strives for proportionality between the offered end-user performance level and the corresponding resource pricing, promoting fairness. The algorithm is validated via full-wave electromagnetic simulations and applies to multiple RIS functionalities and frequency bands.

Traffic categorization is considered of paramount importance in the network security sector, as well as the first stage in network anomaly detection, or in a network-based intrusion detection system (IDS). This paper introduces an artificial intelligence (AI) network traffic classification pipeline, including the employment of state-of-the-art image-based neural network models, namely Vision Transformers (ViT) and Convolutional Neural Networks (CNN), whereas the primary element of this pipeline is the transformation of raw traffic data into grayscale pictures introducing a properly developed IDS-Vision Toolkit as well. This approach extracts characteristics from network traffic data without requiring domain expertise and could be easily adapted to new network protocols and technologies (i.e. 5G). Furthermore, the proposed method was tested on the CIC-IDS-2017 dataset and compared to a well-known feature extraction strategy on the same dataset. Finally, it surpasses all suggested binary classification algorithms for the CIC-IDS-2017 dataset to the best of our knowledge, paving the path for further exploitation in the 5G domain to successfully address related cybersecurity challenges.

Reconfigurable Intelligent Surfaces (RIS) constitute a promising technology that could fulfill the extreme performance and capacity needs of the upcoming 6G wireless networks, by offering software-defined control over wireless propagation phenomena. Despite the existence of many theoretical models describing various aspects of RIS from the signal processing perspective (e.g., channel fading models), there is no open platform to simulate and study their actual physical-layer behavior, especially in the multi-RIS case. In this paper, we develop an open simulation platform, aimed at modeling the physical-layer electromagnetic coupling and propagation between RIS pairs. We present the platform by initially designing a basic unit cell, and then proceeding to progressively model and simulate multiple and larger RISs. The platform can be used for producing verifiable stochastic models for wireless communication in multi-RIS deployments, such as vehicle-to-everything (V2X) communications in autonomous vehicles and cybersecurity schemes, while its code is freely available to the public.

Photonic circuits pave the way to ultrafast computing and real-time inference of applications with paramount importance, such as imaging flow cytometry (IFC). However, current implementations exhibit inherent restrictions that consequently diminish the neural networks (NNs) complexity that can be supported.

Abstract: Security, Trust and Reliability are crucial issues in mobile 5G networks from both hardware and software perspectives. These issues are of significant importance when considering implementations over distributed environments, i.e., corporate Cloud environment over massively virtualized infrastructures as envisioned in the 5G service provision paradigm. The SANCUS1 solution intends providing a modular framework integrating different engines in order to enable next‐generation 5G system networks to perform automated and intelligent analysis of their firmware images at massive scale, as well as the validation of applications and services. SANCUS also proposes a proactive risk assessment of network applications and services by means of maximising the overall system resilience in terms of security, privacy and reliability. This paper presents an overview of the SANCUS architecture in its current release as well as the pilots use cases that will be demonstrated at the end of the project and used for validating the concepts.

Evidence for a causal relationship between air pollution and disease is present in ninety percent of the urban population, having an impact mainly on predisposed individuals. Respiratory health is a condition highly affected by outdoor and indoor air pollution. In more detail, exposure to a variety of environmental air pollutants may exacerbate a group of respiratory diseases comprising by chronic obstructive pulmonary disease, asthma, respiratory infection, and lung cancer. In this paper, we propose a comprehensive methodology to implement an air quality monitoring system towards the prevention of the effects of air pollution on respiratory health. Our approach combines distributed sensors that collect air quality information at regular intervals. Outdoor air quality monitoring stations, indoor air quality devices, and wearables, form the sensor network in the proposed system. The air quality monitoring system includes also an artificial intelligent monitoring module and a clinical decision support module, for personalized guidance, which has been developed and successfully deployed in asthma disease.

Geometry contraction algorithms are introduced in this work to implement the diverse respiratory configurations of lung related diseases associated with airways obstructions. In addition, computational fluid dynamics (CFD) techniques along with fluid particle tracing (FPT) methods are utilized to efficiently evaluate the behavior of the airflow during the inhalation period, as well as to clarify the features of the inhaled particles in terms of regional deposition. Useful deductions are drawn regarding personalized medication in obstructed conditions.