Cyber-Secure Wearable, Ultra-Low-Power Networked Sensors for Soldier Health Monitoring: Insights from the CAPSARII project

Luciano Bozzi, Christian Celidonio, Umberto Nuzzi, Massimo Biagini, Stefano Cherubin, Asbjørn Djupdal, Tor Andre Haugdahl, Andrea Aliverti, Alessandra Angelucci, Giovanni Agosta, Gerardo Pelosi, Paolo Belluco, Samuele Polistina, Gaia Strada, Luigi Sossi Ortega, Alexandra-Ioana Albu, Riccardo Volpi, Luigi Malagò, Michael Schneider, Xabier Eguiluz, Haritz Saiz, María ViqueriaarXiv 2026 · DOI · arXiv

TL;DR

The CAPSARII project is developing a secure, ultra-low-power wearable ecosystem designed to monitor soldier health in real-time within the "Internet of Battlefield Things" (IoBT). By integrating smart textiles, edge-deployed AI on RISC-V microcontrollers, and hybrid post-quantum cryptography, the project aims to provide commanders with actionable insights into pilot fatigue and warfighter stress without compromising data security or battery life.

What Problem Does This Paper Address?

Modern military operations are increasingly volatile, demanding better coordination and improved soldier survivability. While consumer wearables for health tracking are ubiquitous, they fail to meet the "battlefield" standard. Military environments impose extreme constraints: temperatures from -15°C to +55°C, high shock/vibration, and compromised network availability.

Furthermore, data privacy is a matter of life and death. Traditional IoT protocols are often too "chatty" (revealing a soldier's location) or lack the cryptographic muscle to resist sophisticated cyberattacks or future quantum threats. CAPSARII addresses the gap between laboratory-grade physiological monitoring and mission-ready deployment, focusing on two primary use cases: pilot fatigue during long missions and warfighters in harsh environments (extreme heat, humidity, or CBRN threats).

Methods

The authors propose a multi-tiered architecture that moves computation as close to the sensor as possible:

  1. Smart Textiles & Perception: Sensors for heart rate, respiration, muscle activity, and Galvanic Skin Response (GSR) are embedded directly into garments.
  2. Edge AI & HW/SW Co-design: To minimize latency and save power, AI models are deployed on secure RISC-V microcontrollers. The team utilizes a "multifaceted model compression strategy," including:
  3. Pruning: Removing redundant neural connections.
  4. Quantization: Reducing the precision of weights to save memory.
  5. Knowledge Distillation: Training small "student" models to mimic large "teacher" networks.
  6. Precision Tuning: Using the TAFFO framework to optimize floating-point to fixed-point conversions.
  7. The IoBT Stack: Data flows through a Body Area Network (BAN) and Squad Area Network (SAN) mesh. The project explores Processing-in-Memory (PIM) accelerators to handle memory-bound AI tasks with minimal energy draw.
  8. Security: The system acts as a "trust anchor," employing Lightweight Cryptography (LWC) for efficiency and investigating Post-Quantum Cryptography (PQC) to future-proof against quantum adversaries.

Datasets

To establish a baseline for stress classification, the researchers used:

  • Public Stress Detection Dataset (Hongn et al.): Physiological signals from 36 subjects recorded via an Empatica E4 device during stress-inducing tasks like the Stroop Test and Trier Mental Challenge.
  • Berkeley Event-Based Eye-Tracking Dataset: Used for validating neurophysiological monitoring and velocity estimation algorithms for eye-tracking sensors.

Key Insights and Ideas

  • Features over Raw Data: In their Transformer experiments, models trained on extracted features significantly outperformed those trained on raw signals. This suggests that even with advanced deep learning, domain-specific signal processing remains vital for high-accuracy edge deployment.
  • The Mobility Gap: Most traditional IoT protocols (like ZigBee or Thread) are designed for static networks. On the battlefield, nodes move rapidly and networks split/rejoin constantly. The paper identifies a critical need for new discovery mechanisms that don't "spoil" a user's location through excessive signaling.
  • Hybrid Security is Mandatory: PQC offers long-term sustainability but has a massive memory footprint. LWC is efficient but lacks quantum resilience. The "aha moment" here is the strategic imperative for hybrid cryptographic protocols that can run on constrained microcontrollers while satisfying NATO/EU defense standards.

Key Results

The team benchmarked several AI models for stress classification:

  • XGBoost (Tree-Ensemble): Achieved the best overall performance in terms of Accuracy (0.901 ± 0.023) and F1 Score (0.920 ± 0.019).
  • Transformer (Feature-based): Attained the highest AUC (0.963 ± 0.025), making it a strong candidate for nuanced fatigue detection.
  • Energy Efficiency: The project targets missions exceeding 48 hours without a recharge, a goal they aim to meet through the combination of RISC-V efficiency and PIM accelerators.

Limitations and Future Work

The paper concludes with a "gap analysis" that serves as a roadmap for the field:

  • Validation Standards: There is currently no universal standard for physiological signal quality in smart textiles, especially after repeated mechanical stress and laundering.
  • Latency Spikes: While BLE and UWB offer low latency, they can suffer from spikes under high mobility. Future work will investigate protocols like Zenoh to maintain situational awareness in contested environments.
  • PIM Programming: Current precision-tuning tools (like TAFFO) aren't yet fully optimized for the unique programming paradigms required by Processing-in-Memory hardware.
  • Strategic Sovereignty: The authors note that many current wearables are produced outside the EU, highlighting the importance of developing a fully European-integrated IoBT stack.

Summary updated

Abstract

The CAPSARII project advances a cyber-secure, ultra-low-power wearable platform and Internet of Battlefield Things (IoBT) framework to support real-time monitoring of military personnel and data-driven decision support. The architecture integrates smart-textile sensing, a body/squad area mesh, edge-deployed AI on secure RISC-V microcontrollers, and cloud/federated analytics for cross-mission learning and comparative studies. Two representative use cases--pilot fatigue and warfighters operating in harsh environments--guide requirements, validation, and interface design (including HUD and haptic feedback). As an initial step toward embedded models, we benchmark stress classification on a public wearable dataset: tree-ensemble baselines achieve state-of-the-art performance for accuracy and F1 (e.g., XGBoost: $\mathrm{Accuracy}=0.901\!\pm\!0.023$, $\mathrm{F1}=0.920\!\pm\!0.019$), while a feature-based Transformer attains the highest AUC ($0.963\!\pm\!0.025$), informing feature-engineering and compression choices for edge deployment. Building on these results, CAPSARII employs HW/SW co-design, model compression (pruning, quantization, distillation), and precision tuning to meet stringent energy and memory budgets; it further investigates Processing-in-Memory accelerators for future efficiency gains. A structured gap analysis highlights the remaining challenges: establishing validation standards for physiological signal quality and textile durability; achieving low-latency, scalable BAN/SAN communications under mobility and interference; and deploying hybrid lightweight/post-quantum cryptography with robust device identity and side-channel resilience on constrained endpoints. By consolidating these elements into a coherent, secure IoBT stack, CAPSARII delineates a practical roadmap from laboratory baselines to mission-ready soldier monitoring in both training and operational scenarios.
BibTeX
@misc{bozzi2026cyber,
  author = {Luciano Bozzi and Christian Celidonio and Umberto Nuzzi and Massimo Biagini and Stefano Cherubin and Asbjørn Djupdal and Tor Andre Haugdahl and Andrea Aliverti and Alessandra Angelucci and Giovanni Agosta and Gerardo Pelosi and Paolo Belluco and Samuele Polistina and Gaia Strada and Luigi Sossi Ortega and Alexandra-Ioana Albu and Riccardo Volpi and Luigi Malagò and Michael Schneider and Xabier Eguiluz and Haritz Saiz and María Viqueria},
  title = {{Cyber-Secure Wearable, Ultra-Low-Power Networked Sensors for Soldier Health Monitoring: Insights from the CAPSARII project}},
  year = {2026},
  eprint = {2611.20187},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.20187},
}
Figure