http://hdl.handle.net/10900/163069 Sat, 04 Jul 2026 00:09:35 GMT 2026-07-04T00:09:35Z https://publikationen.uni-tuebingen.de:443/xmlui/bitstream/id/92bea732-d8de-47c1-a4dd-c8c71f6d5e4b/ http://hdl.handle.net/10900/163069 http://hdl.handle.net/10900/163784 Scalable Cybersecurity Training: Integrating Virtual and Physical Security Teaching Environments Bechtel, Lukas; Schramm, Markus; Popperl, Lukas; Heer, Tobias The number of cybersecurity incidents increases year over year. Cybersecurity education requires hands-on experience to protect infrastructure and services against hackers. However, existing teaching infrastructures face scalability and hardware integration challenges. This paper presents a semivirtualized security teaching infrastructure combining virtual infrastructure, physical hardware access, and an Attack & Defense framework. The infrastructure is based on a Proxmox cluster, managed through a self-developed platform that allows parallel access to different courses. The teaching concept enables students to solve team-based exercises on personal laptops. Using personal laptops motivates students to create and maintain their own set of tools for cybersecurity analysis. Automated scoring and hardware interaction enhance engagement, providing a flexible platform for practical cybersecurity training. Thu, 03 Apr 2025 00:00:00 GMT http://hdl.handle.net/10900/163784 2025-04-03T00:00:00Z http://hdl.handle.net/10900/163782 Automated Test Bench for High-Performance Network Equipment Steinert, Benjamin; Paradzik, Gabriel; Menth, Michael This paper presents an automated test bench that supports reproducible and holistic benchmarking of data plane and control plane performance. The modular architecture integrates the hardware-based traffic generator P4TG and the software-based traffic generator iperf3 for precise control over test traffic. Additionally, it supports automated Device under Test (DuT) reconfiguration between test runs and metric collection. A case study demonstrates the feasibility of the approach by measuring the performance of a modern P4-based COTS data center switch. Thu, 03 Apr 2025 00:00:00 GMT http://hdl.handle.net/10900/163782 2025-04-03T00:00:00Z http://hdl.handle.net/10900/163772 Dynamic Data Plane Updates using Lua and libmoon Simon, Manuel; Gallenmüller, Sebastian; Carle, Georg Upcoming communication networks, such as 6G, require both high performance and reliability, while service updates typically introduce service downtimes. This study explores dynamic network function updates using libmoon, a DPDK-based high-performance packet processing framework. The approach enables seamless, on-the-fly updates of network functions. By leveraging LuaJIT, we profit from just-in-time (JIT) compilation, allowing for efficient per-flow function updates. Our evaluation demonstrates the feasibility of runtime re-programmability in network data planes. We show the induced latencies of runtime changes and examine cross-flow and cross-core influences. Moreover, we investigate the effects of JIT compilation and show the significance of JIT compilation for long-term performance. Thu, 03 Apr 2025 00:00:00 GMT http://hdl.handle.net/10900/163772 2025-04-03T00:00:00Z http://hdl.handle.net/10900/163769 Enhancing Spatiotemporal Networks with xLSTM: A Scalar LSTM Approach for 5G Traffic Forecasting Bettouche, Zineddine; Ali, Khalid; Fischer, Andreas; Kassler, Andreas Accurate spatiotemporal traffic forecasting is vital for optimizing 5G networks. Traditional LSTM models struggle with capturing complex spatiotemporal dependencies, limiting predictive performance. To address this, we propose an enhanced Spatiotemporal Network (STN) integrating Scalar LSTM (sLSTM), a more efficient variant designed to improve temporal modeling while reducing computational complexity. Our dualpath STN processes the input through an sLSTM for sequential feature extraction and a three-layer Conv3D path for spatial feature learning, with both outputs fused in a dedicated fusion layer for enhanced spatiotemporal representation. By incorporating sLSTM, our model stabilizes gradients, accelerates convergence, and enhances accuracy. Experiments on real-world mobile traffic datasets show a 23% MAE reduction over ConvLSTM, with a 30% improvement on unseen data, demonstrating superior generalization for 5G traffic prediction. Thu, 03 Apr 2025 00:00:00 GMT http://hdl.handle.net/10900/163769 2025-04-03T00:00:00Z