The use of the Internet of Things (IoT) is becoming increasingly widespread in the agricultural sector to improve food productivity in response to a growing global population. Low-power wide area networks (LPWANs), such as LoRaWAN and Sigfox, are considered the most suitable communication technologies for IoT applications in smart agriculture. In rural environments, the large-scale deployment of devices for smart agriculture underscores the need to understand interference and coexistence issues among LPWAN technologies, as intensive sensor deployment is becoming the norm. In this paper, we present an in-depth study on interference analysis of Low Power Wide Area Networks (LPWAN) in the context of smart agriculture. Different deployment scenarios are investigated through Monte Carlo simulations using the SEAMCAT tool. In particular, we analyze a network with a device density of 1,000 devices/km2, corresponding to a total of 78,539 devices distributed over a rural area of 19,408 acres. Finally, the results provide new insights into the simultaneous operation of LoRa and Sigfox in emerging smart agriculture applications.
Passive Optical Networks (PONs) are deployed to deliver high-speed networking to both residential and business customers. However, due to the usage of shared-medium technology, specialized PON hardware is required. PON softwarization aims to decouple PON functionality from specialized hardware, moving execution to general-purpose hardware in commercial servers. Softwarizing the lower PON layers is increasingly computationally expensive the further it moves towards the physical layer. We explore a different route by investigating Ethernet as a suitable replacement candidate on the MAC layer, utilizing widely deployed Ethernet hardware. In our proposed architecture, the Ethernet PHY and MAC layers build the foundation with a softwarized PON MAC on top. While current Ethernet standards are not capable of supporting a physical PON system, the prototypical softwarized MAC demonstrates that software-based packet processing and scheduling in combination with Ethernet hardware is able to support a high-performance PON system. Our implementation provisions and manages connected subscriber instances with a custom in-band control plane protocol. The evaluation results show that our softwarized PON MAC can reach total upstream throughput exceeding 23 Gbit/s for two subscribers on a 25 Gbit/s link. Additionally, the limitations of softwarization are explored by optimizing the complex system, e.g., by varying operating frequencies.
Many emerging Artificial Intelligence (AI) workloads rely on cross-regional distributed training, where Collective Communication Libraries (CCLs), such as AllGather, AllReduce, and AlltoAll, operate across multiple datacenters (DCs) interconnected by wide-area networks (WANs). However, while WAN routing can adapt to traffic dynamics at sub-second to second timescales, CCL logical-topology reoptimization is typically applied far more slowly, often every tens or hundreds of collectives. Due to this intrinsic timescale mismatch, CCL performance (e.g., CCL “completion time”) is affected by bandwidth variability in the WAN, motivating us to investigate the individual influence on performance of WAN reconfiguration and CCL reoptimization. In this work, we study the effect of applying fast WAN path reconfiguration and slow CCL logical-topology reconfiguration in cross-regional AI training. Rather than proposing a new coordination policy, we provide an initial evaluation of how these existing mechanisms behave when used individually or jointly under dynamic WAN conditions. By analyzing their combined operation, our study also offers practical insights for rapidly deployable cross-regional training systems, since it builds directly on mechanisms already supported by modern CCLs and WAN controllers. Using a simulated inter-DC environment, we observe that combined WAN reconfiguration and CCL re-optimization can significantly reduce AllGather completion time compared to applying only WAN or only CCL adjustments. These results highlight the tight coupling between CCL behavior and WAN dynamics and motivate the development of more principled coordination mechanisms for practical deployment of cross-regional CCLs.
Root Cause Analysis in cloud-native environments remains challenging due to the dynamic nature of micro-services and the operational expertise required for diagnosis. While recent approaches advance multimodal Root Cause Analysis through learning-based techniques, limited attention has been given to interactive workflows and real-time operational integration. This demo presents an end-to-end RCA framework combining multimodal analysis with an intent-driven conversational interface for Kubernetes environments. The system integrates real-time log ingestion, root cause ranking based on graph correlations and anomaly signals, and LLM-assisted reasoning to support human-in-the-loop diagnosis and operational control. Through a unified dashboard, users can inspect logs, visualise environment topology, and perform pod-level RCA without requiring Kubernetes expertise or command-line knowledge. The demonstration showcases the framework on a containerised testbed featuring OWASP Juice Shop and PostgreSQL, highlighting practical Root Cause Analysis workflows enabled by conversational interaction and graph-based exploration.
The effectiveness of cyber deception technologies relies on their ability to remain indistinguishable from production systems under adversarial scrutiny. Consequently, this paper presents an empirical study of SSH honeypot detection techniques observed during a high-interaction honeypot challenge deployed in the qualification phase of the European Cyber Week (ECW) Capture-the-Flag (CTF) competition. The challenge was designed to log and analyze the methods used by hackers to distinguish a simulated SSH service from a genuine operational environment. We analyze a dataset of adversarial fingerprinting behaviors, including protocol-level, behavioral, and environmental indicators used for honeypot identification. Based on observed attacker actions, we derive a taxonomy of effective SSH honeypot detection vectors. The results highlight practical weaknesses in deployed deception systems and provide actionable insights for improving the realism, robustness, and operational management of SSH honeypots. These findings support the design of more resilient deception infrastructures and contribute to measurement-based network security and threat intelligence practices.
