Intent-Based Networking (IBN) aims to bridge high-level business goals and low-level network configurations, yet existing solutions lack a unified foundation for intent fulfillment, assurance, and conflict resolution. This paper introduces a knowledge-graph (KG)-based framework that formalizes the relationships among intents, network state, and compliance outcomes. The KG serves as a semantic integration layer supporting three core management tasks: (1) translating and validating intents through ontology-driven reasoning, (2) detecting and resolving conflicts among coexisting intents, and (3) assuring that deployed intents remain satisfied through continuous telemetry validation. Conflict coexistence is modeled as a Maximum (Weighted) Independent Set (MWIS) optimization problem on the KG, enabling explainable selection and scheduling of compatible intents. Representative use cases demonstrate how the approach improves semantic consistency and operational traceability in multi-tenant and cross-domain network environments.

Management for Industry 4.0/5.0 will require considerable engineering effort. To reduce such efforts by cost and time is a considerable challenges. This paper highlight that the usage of Service Oriented Architecture and microservice architecture technologies can create substantial saving on engineering for production automation. The impact on engineering cost and time has been investigated by the Arrowhead Tools project. For the purpose 28 industrial use cases were defined. For each of the 28 use cases a baseline for the engineering cost and time was defined, using for each industrial partner the available technology and competence. Based on use case requirements, a dedicated edge microservice architecture was defined and implemented using the Eclipse Arrowhead architecture and reference implementation, complemented by use case-specific edge microservices and tool chain interoperability. The engineering cost and time for implementing the microservice architecture for the use cases were calculated. Costs, as time savings were then determined for each use case. The resulting median cost and time savings were just above 80%, with a range from 30% to 95%.

Greedy geometric routing is an attractive mechanism for next-generation networks due to its stateless operation and reliance on local coordinate information. In hyperbolically embedded networks, however, greedy routing is supported by latent geometric structures whose role in resilience remains poorly understood. In this paper, we investigate how such hidden structures contribute to the robustness of greedy routing under link failures. We focus on two locally identifiable subgraphs: the Hyperbolic Minimum Spanning Tree (HMST), formed by nearest inward neighbors in the radial hierarchy, and a Second-Order Forest (SOF), defined by the second-nearest inward neighbors. Using extensive simulations across five synthetic network ensembles with varying average degree, and complementary experiments on a hyperbolically embedded Internet AS-level topology, we analyze the distribution of greedy routing load via Edge Load Centrality (ELC). In intact networks, SOF edges exhibit ELC distributions close to those of typical links, while HMST edges carry a disproportionate share of greedy traffic. When HMST edges are removed and greedy routes are recomputed, SOF edges become statistically dominant carriers of routing load, with ELC values significantly exceeding those of the remaining network—an effect observed consistently in both synthetic and AS-level graphs. These results reveal a layered, geometry-induced resilience mechanism: a primary latent backbone (HMST) ensures efficient navigation under normal conditions, while a secondary structure (SOF) remains dormant but is dynamically activated under structural damage. Our findings demonstrate that resilience in greedy routing can emerge from local geometric organization rather than explicit redundancy, offering insights for the design of robust, self-organizing next-generation network architectures.

This dissertation studies how large-scale networks can remain dependable and operationally correct as automation, heterogeneity, and decentralisation become increasingly important. It approaches this challenge through two complementary perspectives, one focusing on closed-loop intent operation and the other on routing structure and robustness under partial visibility. In the Intent-Based Networking track, it develops a knowledge-graph-grounded foundation aligned with YANG schemas and management interfaces that makes time-scoped intents, observed state, and compliance outcomes explicitly representable and traceable, and it builds on this with deterministic validation of intent fulfillment, explainable coordination of concurrently effective intents, and evidence-driven assurance from LTE control-plane signaling. In the decentralised routing track, it shows that hyperbolic geometry-aligned representations can impose strong structure on forwarding behavior, where angularly ordered identifiers yield more ordered and more compressible forwarding patterns under both shortest-path routing and greedy navigation. Finally, it identifies a locally inferable greedy-routing backbone and quantifies its robustness implications through targeted versus random link-failure experiments, highlighting a trade-off between geometry-enabled efficiency and backbone fragility under disruption.

The widespread adoption of Large Language Models (LLMs) has shown that complex systems can be effectively controlled through natural language instructions, without requiring users to have prior knowledge of command syntax or configuration mechanisms. However, in domain-specific applications, accurate context recognition and correct interpretation of implicit user intent remain significant challenges. Misunderstood context often leads to imprecise or unreliable outputs, limiting the applicability of LLM-based systems in professional and industrial environments. This paper presents an overview of methods that support telecommunication-related intent understanding, including domain-specific fine-tuning, integration of external knowledge sources, and multi-agent planning mechanisms. The goal of the analysis is to provide design guidelines for solutions that improve the reliability of LLM-driven systems in complex, domain-specific use cases.

The industrial landscape is undergoing a transformative shift towards Industry 5.0, a paradigm characterized by the convergence of sustainability, digital autonomy, and human-centric design. This article focuses on the adoption, enhancement, and implementation of AI-driven hardware, tools, methodologies, and semiconductor technologies in this progression. We present here a comprehensive strategy from the AIMS5.0 project with the objective of connecting academic developments with practical industrial use, fostering a harmonious relationship between humans and machines to improve efficiency, spur innovation, and enhance adaptability. Hence we show here our global vision, and examples of how the creation of AI-based industrial solutions is supported by novel AI-tool chains, advancements in hardware, and tools supporting human aspects.

As artificial intelligence (AI) regulations evolve and the regulatory landscape develops and becomes be more complex, ensuring compliance with ethical guidelines and legal frameworks remains a challenge for AI developers. This paper introduces an AI-driven self-assessment chatbot designed to assist users in navigating the European Union AI Act and related standards. Leveraging a Retrieval-Augmented Generation (RAG) framework, the chatbot enables real-time, context-aware compliance verification by retrieving relevant regulatory texts and providing tailored guidance. By integrating both public and proprietary standards, it streamlines regulatory adherence, reduces complexity, and fosters responsible AI development. The paper explores the chatbot's architecture, comparing naive and graph-based RAG models, and discusses its potential impact on AI governance.

The convergence of next-generation wireless communication technologies and modern energy infrastructure presents a promising path toward sustainable and intelligent systems. This survey explores how beyond-5G and 6G communication technologies can support the greening of Industrial Internet of Things (IIoT) systems and smart grids. It highlights the critical challenges in achieving energy efficiency, interoperability, and real-time responsiveness across different domains. The paper reviews key enablers such as LPWAN, wake-up radios, mobile edge computing, and energy harvesting techniques for green IoT, as well as optimization strategies for 5G/6G networks and data center operations. Furthermore, it examines the role of 5G in enabling reliable, ultra-low-latency data communication for advanced smart grid applications, such as distributed generation, precise load control, and intelligent feeder automation. Through a structured analysis of recent advances and open research problems, the paper aims to identify essential directions for future research and development in building energy-efficient, resilient, and scalable smart infrastructures powered by intelligent wireless networks.

This paper presents the methodology for using LLMs to ease core network signaling analysis. This is done by applying RAG techniques to process standards that describe protocol data formats - and then asking natural language questions about actual capture traces. Analyzing 5G networks is very challenging due to the complex and dynamic nature of signaling protocols. Unlike previous generations, protocol fields and values are described in a human-readable format, enabling textual post-processing, and the direct application of LLM models. Intent-based network management involves natural language-based human interaction with the networking equipment so the desired outcome is achieved without step-by-step instructions and settings by the human. This paper proposes a novel approach that uses the combination of Retrieval Augmented Generation (RAG) and Langchain to automatically answer human questions regarding signalling data. This toolchain makes the analysis part of network fault management intent-based. Moreover, by training LLMs on a vast corpus of standardized signaling data, we demonstrate the model's ability to generate realistic test data. This approach improves the efficiency of automated test environments, ensuring the reliability and performance of networks in real-world conditions.

Accurate and precise time synchronization is a critical requirement for the operation and management of modern packet-switched networks, particularly in the context of 5G and emerging 6G technologies. This dissertation digest paper explores advanced methods to enhance the precision and accuracy of clock synchronization across diverse network scenarios. The research addresses three key topics divided into four thesis groups, focusing on the challenges posed by unknown propagation delays, queuing delays, and channel variations. While traditional approaches often prioritize accuracy (bias reduction), this work also emphasizes improving precision (variance reduction) through innovative filtering techniques. In particular, the dissertation demonstrates the application adaptive filtering-based methods to optimize clock state estimation. These algorithms are designed to be both protocol- and medium-agnostic, making them applicable to a wide range of network synchronization contexts.

The management of complex Systems of Systems (SoS) often requires interacting with a variety of heterogeneous user interfaces and APIs across different components. This task becomes especially challenging when trying to dynamically manage resources real-time. In this paper, we propose a novel multi-agent architecture for an intent based solver engine, which leverages Large Language Models (LLMs) to facilitate system management via natural language interactions. Our engine utilizes multiple LLM agents to autonomously plan and execute user-intended tasks across various RESTful microservices, such as querying data or modifying system resources. The architecture is designed to support multimodal user input, offering transparency and explainability by allowing users to review and modify plans and workflow steps during execution. This approach combines LLM reasoning capabilities, prompt engineering, and multi-agentic patterns to automate decision-making while ensuring a high level of control for the user. The paper details the architecture and reference implementation of the proposed engine and showcases its capabilities through a practical use case.

The increasing complexity and increasing demands of IT applications, especially in federated multi-cluster environments, pose significant challenges for service orchestration. To address these, Zero-Touch Service Management (ZSM) and intent-based management paradigms are gaining traction, allowing users to specify high-level goals rather than low-level configurations. However, current intent-driven approaches often rely on rigid Domain Specific Languages (DSLs) or graphic user interfaces, limiting expressiveness and usability. In this work, we propose a neurosymbolic intent-based platform that leverages Large Language Models (LLMs) for natural language intent ingestion and Answer Set Programming (ASP), a declarative programming paradigm used for solving complex combinatorial problems. The system translates natural language descriptions of microservice requirements into structured policies, enabling explainable service-to-cluster matching across federated Kubernetes environments. We validate our approach through experiments that evaluate both the syntactic correctness and efficiency of various LLMs in intent translation, as well as the computational time of the symbolic placement algorithm.

Intent-Based Networking (IBN) promises to redefine network management by automating operations to align with high-level user intents. The advent of powerful Generative AI (GenAI) models, including Large Language Models (LLMs), could significantly accelerate this transformation. However, cur-rent research remains narrowly focused on LLM-based intent translation, leaving substantial gaps in understanding how GenAI can be applied across the entire IBN life cycle. This paper aims to bridge these gaps by investigating the wider potential of Generative AI (GenAI) in areas like intent orchestration, moni-toring, compliance assessment, and automated actions. Through a systematic categorization of tasks based on GenAI's suitability and the presentation of a practical use case, this work highlights the critical need for more comprehensive research to fully harness the potential of GenAI in advancing IBN.

The recent worldwide turbulence of events from the pandemic lockdown through increased industrial digitization to geopolitical unease shifted towards new primary targets for the latest generation of DDoS threats. Although certain characteristics of current DDoS attack patterns existed before the pandemic or the cloud platform boom, they have now gained prominence and reached their current level of sophistication. In addition to employing innovative methods and tools, the frequency, scale, and complexity of these attacks have also experienced a significant surge. The amalgamation of diverse attack vectors has paved the way for multi-vector attacks, incorporating a distinctive combination of L3–L7 attacking profiles. The integration of the hit-and-run strategy with the multi-vector approach has notably bolstered the success rate. This paper centers around two main aspects. Firstly, it explores the characteristics of the most recent DDoS attacks identified within actual data center infrastructures. To underscore the changes in attack profiles, we reference samples collected recently from diverse data center networks. Secondly, it offers an extensive overview of the cutting-edge methods and techniques for detecting and mitigating recent attacks. The paper places particular emphasis on the precision and speed of these detection and mitigation approaches, predominantly those related to networking. Additionally, we establish criteria, both quantitative and qualitative, to aid in the development of detection methods capable of addressing the latest threat profiles.

Handling processes for distributed, heterogeneous endpoints with certain centralized decision-making possibilities is a challenge for enterprise services. Many companies already use digital contracting, automated quotation, and purchase order (PO) systems - although following the process and being aware of the status is still troublesome for both the end-users and the company controllers as well. This paper describes a B2B Transaction Platform Architecture concept and realization, fitting to current high-scale enterprise requirements. The foundational elements of this framework consist of atomic events and transactions that combine to form comprehensive business processes. Key elements include activity-based process modeling, immutable node structures, publish/subscribe notification systems, and aggregation mechanisms for unambiguous process-state tracking. Balancing distributed execution with centralized decision-making, this system ensures transactional integrity between stakeholders such as purchasers, suppliers, and financial operators. This architecture aims to handle the competing demands of multiparty transaction coordination, scalability, and centralized administrative control while leveraging the benefits of a distributed cloud infrastructure.

Greedy geometric routing enables scalable, stateless communication in IoT and edge networks by relying solely on local coordinate information. However, this efficiency is underpinned by a structural backbone—the Hyperbolic Minimum Spanning Tree (HMST)—which introduces critical implications for performance and resilience. We analyze the HMST as a locally inferable routing scaffold emerging from hyperbolic embeddings, using Edge Load Centrality (ELC) to assess its importance. We show that HMST edges carry a disproportionate routing load and that removing them causes up to $50 \%$ failure in greedy routing—far exceeding the impact of random removals. While this reveals a vulnerability, the HMST’s local reconstructability also enables autonomous recovery of the network. We discuss the implications for resilient, self-organizing network management and coordinate-aware security.

The recent advancements in network automation, powered by AI and Generative AI, have created new opportunities to automate both network configuration and operations. This paper outlines the main focus areas of my proposed PhD dissertation, which centers on using AI to streamline network configuration and improve operational efficiency. My research aims to develop solutions for automating network configurations through Generative AI, as well as leveraging AI technologies to enhance Intent-Based Networking functionalities. By integrating AI-driven approaches, my work seeks to address key challenges, including automating network configurations with Large Language Models and developing effective anomaly detection techniques in cellular network traffic.

Meetings are a fundamental part of professional environments, yet extracting actionable insights from them has been challenging. This paper presents Summer, a multi-agent system that processes meeting transcripts using Large Language Models (LLMs) to generate structured, context-aware sum-maries. Summer integrates with project management platforms, enabling automated task extraction and role-specific summaries. The system employs a structured workflow, ensuring high-fidelity information preservation while maintaining reliability and minimizing hallucinations. The evaluation metrics presented and used go beyond the traditional summarization metrics, as they take into account how well critical information survives the trans-formation process. The validation of the system is demonstrated through processing transcripts of network management meetings, showing the capabilities of controlled LLM workflows for real-world business applications.

Engineering tools support the process of creating, operating, maintaining, and evolving systems throughout their lifecycle. Toolchains are sequences of tools that build on each others’ output during this procedure. The complete chain of tools itself may not even be recognized by the humans who utilize them, people may just recognize the right tool being used at the right place in time. Modern engineering processes, however, do not value such ad-hoc choice of tooling, because of their uncontrolled nature. Building upon the Extended Automation Engineering Model defined by the IEC 81346 standard, this paper proposes to automate the toolchain building and execution process for Cyber-Physical System of Systems (CPSoS), utilizing key principles of the Eclipse Arrowhead framework. The proposed toolchain automation solution addresses issues such as tool interoperability, interaction, automation, and dynamic choreography. The feasibility of this set of integrated concepts is validated through an Arrowhead-based toolchain choreography demonstration. Note to Practitioners —The paper discusses approaches to the automated execution of various industry-related processes. As the processes are becoming more complex and involve numerous systems which have to be orchestrated, a simple and preprogrammed workflow is not enough anymore. Therefore, building on top of the principles of the Eclipse Arrowhead framework, an adequate model of toolchains, allowing for their automated execution, is proposed. Different approaches to supervision of toolchain execution are discussed showing the benefits of reaching higher automation levels. Further, four adoption levels are introduced, which are a measure of the toolchain automation progress. Finally, a simplified demonstrator is shown and steps to elevate it to higher adoption levels are highlighted. To ensure that the approach is industry-oriented, several examples of how the proposed methodology can be used in the industrial context are discussed.

The adoption of AI in industrial applications has progressed slower than anticipated, primarily due to the absence of best practices and standardized guidelines for efficient implementation. To address this, the Artificial Intelligence in Manufacturing leading to Sustainability and Industry 5.0 (AIMS5.0) project introduces 20 use cases from globally renowned companies, offering a comprehensive opportunity to analyze the challenges of AI implementation in industrial contexts. This paper presents an empirical study aimed at identifying and validating the key challenges of AI adoption across various industrial sectors, establishing a ranking to highlight the most pressing issues. Based on these findings, the AIMS5.0 AI Toolbox will target the most critical challenges and provide guidance throughout the entire AI adoption lifecycle. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)