Three papers accepted at IEEE CCNC 2027

Three papers from NSLab have been accepted at the IEEE Consumer Communications & Networking Conference (CCNC 2027), to be held in Las Vegas in January 2027.

A Lightweight Hybrid QVAE for Passive Real-Time Anomaly Detection in IIoT/OT Networks (M. Bakro, A. Marotta, W. Tiberti, M. Beseda, O. Odoardi, I. Salvatore, P. Di Marco) presents a hybrid quantum variational autoencoder that monitors industrial network traffic passively, one second at a time, and flags anomalies with a mean inference latency of about 3 ms. The framework was evaluated on a real industrial testbed, in collaboration with S.EL.ME.C. SRL.

Towards Standardized Energy Profiling for the Kubernetes Edge: Reproducible Campaigns with Quality-Labeled Power Sources (C. Centofanti, A. Marotta, A. D'Errico, F. Graziosi) addresses a prerequisite of energy-aware orchestration: trustworthy per-node power data on the hardware that actually populates edge sites, where standard software power meters are unavailable or out of their calibration domain. The paper presents an open framework for reproducible energy profiling campaigns on Kubernetes clusters and releases all artifacts.

Orchestrating Split AI Computation in the Client-AI Era (A. Marotta, C. Centofanti, M. Zenadocchio, F. Graziosi) revisits split computing now that consumer clients ship with dedicated inference accelerators. By metering both sides of a client–edge split across three client classes, including an NPU-equipped AI PC, the paper shows that whenever the client has an accelerator, local inference wins on system-wide energy and latency, and offloading mostly displaces energy onto the edge node.

The second and third papers are the result of the ongoing collaboration between NSLab and NexTrainE, the University of L'Aquila spin-off.