https://hdl.handle.net/20.500.12761/2 2026-07-15T06:00:07Z https://hdl.handle.net/20.500.12761/2051 Disentangling Simultaneous Spatiotemporal Events in Mobile Network Traffic Data: The Case of the 2025 UEFA Champions League Final Pirri, Máximo; Madariaga, Diego; Smoreda, Zbigniew; Fiore, Marco Large-scale social gatherings often generate distinctive patterns of activity in mobile networks due to the concentration of a substantial amount of users. While prior work has shown that specific individual events leave clear identifiable signatures in aggregate network traffic, disentangling the effects generated on mobile data usage by concurrent and co-located happenings entails significant added complexity. In this paper, we present a methodology to detect different but coinciding mass manifestations by unraveling distinct traffic signatures associated with the coexisting behaviors, thus allowing to monitor the spatiotemporal evolution of each event in isolation. We demonstrate the effectiveness of our approach in a practical use case, i.e. the combination of cheerful celebrations and social unrest episodes that accompanied Paris Saint-Germain’s 2025 UEFA Champions League victory in the city of Paris, France. Using mobile network traffic measurements collected by a leading network operator, we successfully separate the mobile traffic consumption patterns of peaceful partying crowds from those of rioters that confronted local police forces, and reconstruct the directional flows of the different groups across the city. 2026-06-01T00:00:00Z https://hdl.handle.net/20.500.12761/2050 Demo: ISAC Real-Time Experimentation Platform Ruiz, Rafael; Canil, Marco; Lacruz, Jesús Omar; Widmer, Joerg Integrated Sensing And Communication (ISAC) is a key enabler for future wireless systems, yet realizing it in practice requires executing tightly coupled communication and sensing pipelines under strict real-time constraints. In particular, maintaining a continuous high-rate communication link while extracting sensing features introduces fundamental challenges in computation partitioning and latency management across hardware platforms. In this demonstration, we present a real-time Millimeter-Wave (mmWave) ISAC system which exposes and addresses these challenges through a cross-layer hardware/software design. Our approach preserves a standard 5G-like communication pipeline and enables sensing by appending bursts of reference signals for beam-sweeping-based channel probing, requiring only minimal modifications to the payload structure. The system partitions the processing across the FPGA logic, an embedded ARM processor, and an external host, enabling synchronous transmission, real-time channel extraction, and continuous sensing without disrupting the communication link. We demonstrate an end-to-end system where Channel State Information (CSI) is extracted on-device and streamed for realtime processing and visualization of micro-Doppler signatures from moving targets. 2026-06-20T00:00:00Z https://hdl.handle.net/20.500.12761/2049 Setchain algorithms for blockchain scalability Karmegam, Arivarasan; Luz Bianchi, Gabina; Capretto, Margarita; Ceresa, Martín; Fernández Anta, Antonio; Sánchez, César Blockchain scalability remains a longstanding obstacle to its broader adoption. To mitigate this limitation, numerous approaches have been proposed to improve blockchain throughput and efficiency. One such approach is Setchain, a reliable distributed object that improves scalability by relaxing the strict total-order requirement among transactions. Setchain arranges elements as an ordered sequence of sets, called epochs. Elements inside an epoch have no internal order, while epochs themselves are totally ordered. In this work, we introduce and evaluate three Setchain algorithms built on top of a block-based ledger. Vanilla is a straightforward baseline implementation. Compresschain groups elements into batches and compresses them before appending them as epochs on the ledger. Hashchain instead maps each batch to a fixed-length hash, which is then appended as an epoch in the ledger. This design requires Hashchain to rely on a distributed service to retrieve the batch contents corresponding to a given hash. To enable light clients to safely interact with a single server, the proposed algorithms store epoch-proofs within the Setchain. An epoch-proof is the hash of the epoch, cryptographically signed by a server. A client can verify the correctness of an epoch with epoch-proofs, where f is the maximum number of Byzantine servers assumed. We implement all three Setchain variants on the CometBFT blockchain application platform and deploy them as Docker-based nodes in a cluster. We evaluate performance under multiple settings using clusters with four, seven, and ten servers. The results indicate that the Setchain algorithms achieve throughput orders of magnitude higher than that of the underlying blockchain, while reaching finality with latency under 4 s. 2026-06-28T00:00:00Z https://hdl.handle.net/20.500.12761/2048 Index-only Backdoor Vetting for Secure Federated PEFT of Large Language Models Dogani, Javad; İşler, Devriş; Laoutaris, Nikolaos Federated parameter-efficient fine-tuning (PEFT) enables customizing large language models on private data, yet it is vulnerable to backdoor poisoning—especially when privacy constraints prevent inspection of per-client real-valued updates. We exploit the intuition that poisoning leaves a similar backdoor imprint in which adapter coordinates become salient, so overlap in salient-index supports remains informative even without values. We introduce INDEXGUARD, an unsupervised index-only vetting primitive in which clients send only Top-$K$ salient update indices and the server operates on the induced overlap geometry, clustering clients and filtering cohesion-outlier groups before aggregation. We analyze support stability under bounded rescaling and separability under shared-trigger poisoning under non-IID drift. Across attacks, backbones, and PEFT variants, INDEXGUARD provides end-to-end mitigation, preserving clean accuracy while achieving performance comparable to centralized methods. 2026-07-06T00:00:00Z