CSAN 2026

Session 1. Industry 4.0 and embedded wireless sensor/actuator systems

sciforum-177129 Experimental Validation of the BlueSkyTec Post-Quantum Cybersecurity Defence Mechanism

sciforum-182702 A Bibliometric and Thematic Analysis of Hybrid AI for Predictive Maintenance and Prognostics & Health Management in Cyber-Physical Manufacturing Systems (2018–2025)

Session 3. WSAN and next-generation networks (5G, 6G, etc.)

sciforum-183583 OpenWrt-Oriented Cooperative Multi-AP Wi-Fi Sensing with Reliability-Aware Fusion and QoS-Bounded Measurement Control

Session 4. Applications of WSAN in agriculture, vehicle, wearable sensors, smart cities, manufacturing, mobile systems, and health and medical care

sciforum-177292 Privacy-Preserving Multiparty Computation for Quantum-Resilient Healthcare Sensor Networks A Systematic Review

Session 5. Big Data, Computing and Artificial Intelligence

sciforum-186393 Topological Data Analysis (TDA)-based Feature Engineering for Reinforcement Learning-based Trading Strategies in Financial Markets

sciforum-183642 Heuristic-Based Detection of Anomalous Behavior in Software-Defined Networks

Mathematical Artificial Intelligence for Networking and Communications

Prof. Dr. Shui Yu

University of Technology Sydney, Sydney, Australia

Abstract

The application of AI is extremely hot today. However, the theoretical part is quite naïve, which causes a big problem in killer applications, such as cloud, networking, and communications. In this talk, we firstly introduce the landscape of the current status of theoretical effort in AI. Secondly, we present the related popular mathematical tools based on our study. Finally, we show how the tools are used and could be used in the practice of networking and communications. We hope this talk will shed light to interested people to explore the uncharted land together.

Advancing AI and Robotics: Innovations in AgTech, Intelligent Transportation, and Assistive Technologies

Prof. Dr. Hovannes Kulhandjian

California State University, USA

Abstract

Autonomous systems are rapidly moving from lab prototypes to real deployments across farms, cities, and community spaces. This talk surveys practical AI-robotics pipelines—perception, planning, control, and human-in-the-loop design—that my group has translated into working field systems. In AgTech, we’ll discuss modular platforms for fruit harvesting, precision spraying, and pollination, combining edge vision (lightweight detectors on Jetson), robust actuation (CoreXY mechanisms, 6-DoF arms), and data-driven operations. In civil infrastructure and transportation, we will focus on AI-powered bridge and roadway inspection—vision/LiDAR-based defect detection, surface-distress mapping, and structural-condition scoring—as well as road-user safety through vehicle, cyclist, and pedestrian detection with near-miss analytics for proactive hazard mitigation. For assistive technologies, we’ll cover CrossBot, a smart crossing-assist robot, and ongoing work integrating low-visibility sensing (IR/radar/audio fusion) and brain–computer interface (BCI) concepts to expand accessibility. This includes results stemming from my U.S. patent on low-visibility human/animal detection using multimodal sensing, and lessons learned when hardening research prototypes for real environments. Across these domains, I’ll share insights on system architecture, model selection (from classical computer vision to compact deep nets), domain shift and on-device adaptation, and workforce development through student-centered field deployments—closing with an outlook on scalable, standards-aligned deployments and academia–industry–community collaboration.

Multimodal Explainable Artificial Intelligence (XAI) Applied to Parkinson’s and Alzheimer’s Diseases

Dr. Anastasia Bougea

National Technical University of Athens, Greece

Abstract

Neurodegenerative disorders, specifically Parkinson’s (PD) and Alzheimer’s disease (AD), present significant diagnostic challenges due to their heterogeneous symptomatology and overlapping clinical features. While deep learning has demonstrated remarkable potential in automated diagnosis, the "black box" nature of these models hinders their integration into clinical workflows, where interpretability is paramount. This presentation introduces a comprehensive framework for Multimodal Explainable Artificial Intelligence (XAI) tailored to the early detection and progression monitoring of PD and AD.

By fusing diverse data modalities—such as neuroimaging (MRI/PET), electrophysiological signals (EEG), and non-invasive biomarkers (voice or gait analysis)—the proposed approach captures the multifaceted pathology of these diseases more effectively than unimodal systems. Crucially, the framework integrates XAI techniques, including SHAP (SHapley Additive exPlanations) and attention mapping, to elucidate the algorithmic decision-making process. This allows clinicians to visualize specific anatomical regions or signal features driving the model’s predictions. The presented findings demonstrate that multimodal fusion not only enhances diagnostic accuracy but also provides transparent, medically relevant insights. Ultimately, this work bridges the gap between high-performance AI and clinical trust, paving the way for reliable, personalized computer-aided diagnostic tools in neurology.

ForestWatch: Observing, Responding and Predicting the Deforestation Events using Wireless Sensors

Dr. Masood Ur Rehman

University of Glasgow, UK

Abstract

The world today is striving to be smarter through the influx of innovative technologies. Communication and sensing devices/systems are a vital part of these technologies with applications beyond their usual domain and playing a crucial role in every aspect of our lives including healthcare, automation, transport, weather, education, entertainment, and security. Uninhibited industrialization, urbanisation and fulfilment of an over-populated planet has brought upon us a climate emergency. Deforestation through depletion of the tree crown cover and degradation of tree health caused by over-exploitation, repeated fires, or diseases is reckoned as one of the major drivers behind catastrophic climate changes. It also inflicts long-term economic losses due to limited natural resources, soil erosion, increased flooding, drought, and a domino of unfavourable effects in terms of ecosystem and public health. This talk unfolds the potentials of using a low-cost, autonomous, real-time monitoring and response system utilising intelligent wireless sensor networks and holistic implementation of Internet-of-Things to combat deforestation and avoid a fast-looming climatical calamity.

Security of Critical Infrastructure Cyber-Physical Systems (CPS)


Dr. Qasem Abu Al-Haija

Jordan University of Science and Technology (JUST), Irbid, Jordan

Abstract

Critical infrastructure systems—such as power grids, transportation networks, water treatment facilities, and industrial automation platforms—are increasingly interconnected, automated, and data-driven. This evolution has transformed them into complex cyber-physical systems (CPS) with tightly coupled digital and physical components. While this integration enhances efficiency and real-time control, it also exposes critical assets to new classes of cyber threats capable of causing systemic disruption, physical damage, or large-scale service failures. This talk presents a CPS-oriented perspective on securing critical infrastructure, emphasizing the unique risks that emerge from sensing, actuation, communication, and control loops. I will discuss recent attack patterns—including sensor spoofing, command injection, and coordinated cyber-physical attacks—along with the cascading effects these threats can trigger across dependent subsystems. The presentation outlines a structured methodology for threat modeling in CPS, highlights detection and response challenges, and showcases practical strategies leveraging AI-driven analytics, resilient control design, and intrusion-tolerant architectures. The session aims to provide a holistic understanding of how modern critical infrastructures can be hardened against next-generation cyber-physical adversaries while maintaining safety, reliability, and mission-critical performance.

Bandwidth Conscious Interference Mitigation in D2D Underlay 5G Networks


Dr. Ghazanfar Ali Safdar

School of computing, Engineering and Creative Industries, University of Bedfordshire, Luton, UK

Abstract

Device-to-device (D2D) communication in the cellular band promises a practical approach to increase capacity, reduce latency and energy consumption, and enable new local services without proportionally expanding spectrum or infrastructure. D2D allows direct communication between proximate user equipment’s and achieves three apparent benefits: proximity gain from reduced path loss and lower transmit power; spatial reuse (reuse gain) from allowing multiple short range links to occupy spectrum that would otherwise be centrally scheduled; and hop gain by replacing two hop cellular exchanges with single hop links. These impacts together can significantly boost spectral efficiency and area throughput, reduce end‑to‑end delay for latency sensitive services, and enable functions such as local content distribution, device relaying for cell edge enhancement, and low latency V2X-style services. The capacity for operators to release local traffic from the core also provides operational advantages: lower backhaul and core load, reduced energy expenditure at the network edge, and opportunities to introduce operator controlled proximity services that are both revenue generating and safety critical. This talk presents Multicell Uplink Resource Shared Interference Mitigation Scheme (MURSIMS), a unified framework which jointly considers user mobility, resource (bandwidth) allocation, and transmit‑power regulation to enable reliable in‑band D2D reuse. Results show that MURSIMS delivers substantial improvements in aggregate throughput, spectrum efficiency, and interference containment when compared against other state‑of‑the‑art.

From Sensing to Action: Foundation Models, Edge Intelligence, and Agentic AI for Autonomous Industrial Systems


Dr. Tapiwa Chiwewe

Council for Scientific and Industrial Research, South Africa

Abstract

Industry 4.0 has delivered an abundance of sensing (cameras, accelerometers, acoustic, and thermal sensors now blanket production lines) yet most of this data still flows into dashboards that wait for a human to notice. This keynote presents that we are at the threshold of a fundamental architectural shift: moving from passive networks that report to intelligent sensor-actuator networks that perceive, reason, and act autonomously.

The talk develops this argument across three distinct layers. At the perception layer, specialized, lightweight models distilled from large pretrained networks run directly on embedded devices. We will look at how they detect micro-defects like crack propagation in real time within milliwatt power budgets, illustrated through electroluminescence-based solar panel inspection and vibration-based condition monitoring. At the knowledge layer, foundation models emerge as the solution to industrial AI's chronic label scarcity by acting as teachers for edge models, enabling few-shot adaptation to new products, and transferring knowledge across different machinery and plants. At the action layer, agentic AI closes the loop by correlating multimodal sensor evidence with process parameters to decide when to reject a part, adjust an actuator, or escalate to a human.

Finally, the keynote examines what this autonomous loop makes possible, what it breaks (including hard questions of trust, uncertainty, and safe actuation) and the open research challenges that will define embedded intelligent systems over the next five years.

Enhancing Cybersecurity Resilience in Operational Technology for the Energy Sector: The Role of AI-Driven Anomaly Detection


Prof. Dr. Mohammad Hammoudeh

King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia

Abstract

Artificial intelligence now sits at the centre of both cyber defence and cyber offence, raising a deceptively simple question: is AI a friend, a foe, or a vulnerability? This talk argues that it is all three at once, and that this triple identity is precisely what makes securing and governing it so difficult. The same capabilities that enable real-time intrusion detection, behavioural analytics, and automated malware classification also empower adversaries to automate social engineering, accelerate vulnerability discovery, and mount adversarial attacks—capability, in short, is symmetric across attacker and defender, and only intent differs. Building on this framing, the talk introduces a third, frequently neglected dimension: AI as an attack surface in its own right, exposed to model poisoning, adversarial inputs, and supply-chain risks inherited from open-source and pretrained models. It then turns to the energy sector and operational technology (OT), where the consequences of failure are physical rather than merely informational, examining why the IT/OT convergence has dissolved the historical air gap, why AI-driven anomaly detection developed for IT does not transfer cleanly to OT, and why the scarcity of realistic, shareable OT datasets remains a fundamental barrier to progress. The session closes by making the case for genuine multidisciplinary collaboration and by looking ahead to AI-versus-AI conflict, self-defending systems, and the integration of quantum-resistant cryptography, with a central call to action: to deliberately tip the balance of AI toward defender through cybersecurity-aware architectures, representative datasets, and sound governance.

Multi-Modal Human Activity Recognition: Federated Learning Across Heterogeneous Physiological and Wireless Signal


Asst. Prof. Ng Pai Chet

Singapore Institute of Technology (SIT), Singapore

Abstract

Human Activity Recognition (HAR) has become a cornerstone of modern healthcare, smart environments, and human-computer interaction. While wearable physiological sensors offer deep insights into a user's internal state, ambient wireless sensing provides a non-invasive look at spatial movement. However, combining these modalities raises severe privacy concerns, as continuous monitoring exposes sensitive personal routines and health data. This talk explores the integration of wearable and wireless sensing within a Federated Learning (FL) framework. By allowing decentralized devices to collaboratively train a shared HAR model without exposing raw sensor data, FL mitigates privacy risks. We will discuss the core technical challenges of this paradigm, including data heterogeneity across different users, the synchronization of heterogeneous signals, and communication efficiency. Finally, we will highlight future directions for building robust, multi-modal, and privacy-preserving sensing systems.

Zero-Trust Architecture: Redefining Network Security for Industrial Cyber-Physical Systems


Dr. Daniel Ramotsoela

University of Cape Town (UCT), Cape Town, South Africa

Abstract

The fourth industrial revolution has permanently bridged the digital and physical worlds, transforming isolated industrial control systems into highly interconnected industrial cyber-physical systems (CPS). While this shift shatters traditional network perimeters and drives massive gains in operational agility, it also exponentially expands the attack surface. Traditional perimeter-based security models are no longer sufficient to protect these highly interconnected, heterogeneous environments.

Securing next-generation infrastructure requires a fundamental paradigm shift: Zero-Trust Architecture (ZTA). Grounded in the core principle of 'never trust, always verify,' ZTA eliminates implicit trust based on network location by enforcing continuous device authentication, strict micro-segmentation, and dynamic, context-aware access control policies. However, this paradigm introduces distinct systems engineering challenges. This presentation unpacks the evolving security realities of modern infrastructure and explores the critical tension between deploying robust Zero-Trust protocols and maintaining real-time operational determinism in resource-constrained, time-critical physical environments.

Connect with the Sensors Community In-Person!

We are pleased to announce that the 3rd International Conference on AI Sensors and Transducers (AIS 2026) would be held in Jeju, Korea, from 2–7 August 2026.

Don't miss this chance to showcase your research at this global platform, where you have a chance to present your work to award-winning innovators and network with pioneering experts.

Ms. Sylvie Shan
Ms. Ann Li
Email: csan2026@mdpi.com

For inquiries regarding submissions and sponsorship opportunities, please feel free to contact us.