The deployment of machine learning (ML) on edge devices and ultra-low-power microcontrollers (MCUs) has led to the emergence of two dominant paradigms: Edge ML and TinyML. Edge ML leverages moderately powerful devices—such as embedded systems and AI-capable computing platforms—for real-time inference. In contrast, TinyML pushes the boundaries by enabling ML capabilities on ultra-low-power microcontrollers operating at milliwatt-level power consumption. This paper presents a comprehensive comparative analysis of these two approaches, supported by experimental results from an object detection application (egg counting) implemented on both an MCU-based device (Arduino Nicla Vision) using TinyML and a single-board unit (Raspberry Pi) employing Edge ML. We evaluate key metrics, including power consumption, latency, and the trade-offs between INT8 and FLOAT32 model representations. Our findings show that TinyML achieves significantly lower energy consumption than Edge ML for comparable tasks, positioning it as a more suitable choice for battery-operated IoT applications.

In today’s digital age, access to the Internet is essential, yet a significant digital divide exists, particularly in rural areas of developing nations. This paper presents a Delay Tolerant Networking (DTN) framework that utilizes informal public transportation systems, such as minibus taxis, as mobile data mules to enhance connectivity in these underserved regions. We develop a probabilistic model to capture the randomness in vehicle mobility, including travel times and contact durations at bus stops. Key performance metrics are analyzed, including average data transmission rate and Peak Age of Information (PAoI), to assess the effectiveness of the proposed system. An analytical approximation for the Mean PAoI (MPAoI) is derived and validated through simulations. Case studies from real-world datasets in Nouakchott, Accra, and Addis Ababa demonstrate the practical applicability and scalability of our framework. The findings indicate that leveraging existing transportation networks can significantly bridge the digital divide by providing reliable internet-like connectivity to remote areas.

ABSTRACT The increasing use of the Internet of Medical Things (IoMT) in healthcare highlights privacy and security concerns surrounding sensitive health data. This research focuses on enhancing the security and usability of IoMT for young users through a robust, adaptive continuous authentication model using physiological biometrics on Android devices and heart rate data from smartwatches. By integrating user behavior, environmental context, and health conditions, the model dynamically determines risk, trust, and authorization decisions. Machine learning techniques analyse data related to devices, networks, locations, and user habits while considering demographics like age and medical conditions to assign suitable authenticators. The model balances accuracy and usability, favouring correct positive predictions, but faces limitations such as class imbalance, feature selection, and overfitting, with a false rejection rate (FRR) of 19%. Behavioral biometrics, personalized authentication, and continuous authentication enhance security and accessibility. However, moderate sensitivity affects its ability to capture all positive cases. Age‐group analysis reveals varying engagement with technology, emphasising tailored authentication flows. Future work will explore explainable AI, context‐aware analytics, and advanced risk assessments, integrating complementary smartwatch data like step count for improved accuracy. This research demonstrates the potential of risk‐based adaptive authentication to deliver secure, user‐friendly solutions in complex healthcare environments.

The significance of distributed learning and inference algorithms in Internet of Things (IoT) network is growing since they flexibly distribute computation load between IoT devices and the infrastructure, enhance data privacy, and minimize latency. However, a notable challenge stems from the influence of communication channel conditions on their performance. In this work, we introduce COMSPLIT: a novel communication-aware design for split learning (SL) and inference paradigm tailored to processing time series data in IoT networks. COMSPLIT provides a versatile framework for deploying adaptable SL in IoT networks affected by diverse channel conditions. In conjunction with the integration of an early-exit strategy, and addressing IoT scenarios containing devices with heterogeneous computational capabilities, COMSPLIT represents a comprehensive design solution for communication-aware SL in IoT networks. Numerical results show superior performance of COMSPLIT compared to vanilla SL approaches (that assume ideal communication channel), demonstrating its ability to offer both design simplicity and adaptability to different channel conditions.

As context-aware location-based services (LBS) become increasingly important in many Internet of Things (IoT) verticals, such as logistics or industry 4.0, indoor localization is now an essential feature to be integrated in these solutions. For this purpose, fingerprinting-based solutions arise as a feasible solution, especially when integrating artificial intelligence on the edge, supported by computational and memory-restricted embedded devices, as it does not depend on a cloud-based deployment. In this work, we integrate this new paradigm, known as TinyML, and compare the implementation of a machine learning (ML) model when using only WiFi Received Signal Strength Indicator (RSSI) or WiFi Channel State Information (CSI) data. We tested two different scenarios, a single sample or time series, with different configurations of the trained neural network. Our results show that a CSI data ML model always outperforms an equivalent RSSI approach, with a massive difference in performance for the time-series case.

Cosmic rays (CR) and cosmic ray ensembles (CRE) can reveal new insights about the universe. Scientists need low-cost devices that can detect and record CRE events over large areas. We built a small, cheap, and energy-efficient device that can do this using a TinyML model to identify three types of cosmic ray traces: Worm, Spot, and Line. The model has been successfully deployed in hardware and has been tested using three test-bench scenarios. The results show a successful detection and identification for all cosmic ray traces. The detected trace information can be then sent on the Internet. Our device can help expand the cosmic ray observatory network and can be used for space missions and IoT applications.

Efficient long-range communication is critical for environmental monitoring, especially when dealing with large data transfers in remote areas. We present an AI-driven dynamic RF configuration mode, which combines the advanced capabilities of a novel AI*LoRa model with an enhanced RF configuration request mechanism. By leveraging TinyML, AI*LoRa dynamically adjusts key parameters of the physical layer, based on real-time environmental data, ensuring robust and energy-efficient communication. We conducted extensive real-world testing across distances ranging from a few meters to over 100 kilometers to collect the dataset necessary for training our model. The results demonstrate that our approach achieved over 90% accuracy in predicting optimal settings, leading to an average improvement of over 170% in communication efficiency. These findings underscore AI*LoRa's significant potential to enhance long-range IoT deployments.

Cellular communication standards have been established to ensure connectivity across most urban environments, complemented by deployment hardware and facilities tailored for city life. At the same time, numerous initiatives seek to broaden connectivity to rural and developing areas. However, with nearly half the global population still offline, there is an urgent need to drive research toward enhancing connectivity in areas and conditions that deviate from the norm. This article delves into innovative communication solutions not only for hard-to-reach and extreme environments but also introduces “hard-to-serve” areas as a crucial, yet underexplored, category within the broader spectrum of connectivity challenges.We explore the latest advancements in communication systems designed for environments subject to extreme temperatures, harsh weather, excessive dust, or even disasters such as fires. Our exploration spans the entire communication stack, covering communications on isolated islands, sparsely populated regions, mountainous terrains, and even underwater and underground settings. We highlight system architectures, hardware, materials, algorithms, and other pivotal technologies that promise to connect these challenging areas. Through case studies, we explore the application of 5G for innovative research, long range (LoRa) for audio messages and emails, LoRa wireless connections, free-space optics, communications in underwater and underground scenarios, delay-tolerant networks, satellite links, and the strategic use of shared spectrum and TV white space (TVWS) to improve mobile connectivity in secluded and remote regions. These studies also touch on prevalent challenges such as power outages, regulatory gaps, technological availability, and human resource constraints, where we introduce the concept of peri-urban hard-to-serve areas where populations might struggle with affordability or lack the skills for traditional connectivity solutions. This article provides an exhaustive summary of our research, showcasing how 6G and future networks will play a crucial role in delivering connectivity to areas that are hard-to-reach, hard-to-serve, or subject to extreme conditions (ECs).

The proliferation of Internet of Things (IoT) technologies in various industrial sectors has brought forth new challenges that demand attention for achieving technological maturity. One such challenge is the lack of tools for emulating the diverse components present in IoT architectures, leading to the continuous verification of each component in the chain, which proves to be a complex task. This paper addresses the verification problem in Edge/Fog IoT platforms through comprehensive end-to-end testing. To tackle this challenge, we have developed a modular IoT simulator (SimIoT) capable of efficiently emulating thousands of IoT devices in realistic scenarios, including hospitals, airports, and smart cities. The simulator allows testing of Edge platforms without the need for programming expertise. Furthermore, we demonstrate the feasibility of our simulator by presenting a use case involving the profiling of an open-source IoT platform.

LoRaWAN has proven helpful in many application areas, such as meteorological monitoring in sparsely populated areas. LoRaWAN End Nodes send data to a LoRaWAN Gateway, conveying it to network and application servers using an IP connection. In remote areas, connectivity is often unavailable, but these areas can provide a wealth of data valid for disaster prevention, wild animal tracking, climate assessment, and modeling, among others. We present a LoRaWAN Range Extender, which extends the coverage of existing gateways without requiring any modification to the existing LoRaWAN infrastructure. This device could double the achievable range of a standard LoRaWAN link or enable nodes to be serviced behind obstacles that prevent direct communication with the Gateway. The Range Extender must be within the communication reach of an existing LoRaWAN Gateway while also being reachable by the End Nodes of interest. It can be located in places without cellular coverage, and since it is cheaper to operate and install, it can lower both capital and operational expenses.

The agricultural sector faces several challenges in its efforts to increase its production. One of the major challenges is diseases and insect pests that destroy plants and lead to a decrease in production. The paper addresses precision agriculture and presents the most recent related work on techniques for early detection of plant diseases and insects. Specifically, the paper presents two classes of techniques: first, The application of Image Processing Techniques associated with Machine Learning Algorithms, and second the application of deep learning, in disease detection and recognition. These techniques are discussed and compared. As they require powerful computing equipment that requires a constant power supply and a high bandwidth for their implementation, we propose a new solution more accessible to developing countries, which is based on Tiny Machine Learning (TinyMl), the emerging technology of embedded Machine Learning. Although the TinyMl is running Machine Learning on cheap and low energy devices, it can infer from the models obtained from the machine learning algorithms. The idea is to get the model after training, and then convert it into a lite model with a size that can be inserted into devices with external memory. Implementation issues are discussed.

An Analysis of many IoT deployments showed that most of them can address the sustainable development goals (SDGs) and the UN’s 2030 agenda. Interestingly, most of these projects concentrate on five SDGs: 1) industry, innovation, infrastructure; 2) smart cities and communities; 3) affordable and clean energy; 4) good health and well-being; and 5) responsible production and consumption. Examples include a remote water-monitoring solution that ensures clean water in regions with an indigenous population and smart lighting initiatives in Chinese cities that halve total power output.

Over the last few years we have witnessed an exponential growth in the adoption of LoRaWAN as LPWAN technology for IoT. While LoRaWAN offers many advantages, one of its limitations is the paltry data rate. Most IoT applications don't require a high throughput but there are some that would benefit from a higher data rate. In this paper, we present TurboLoRa, a system that combines the strengths of LoRaWAN while providing a higher data rate by synchronizing the transmission of multiple LoRaWAN devices. Our proposal allows to combine cheap devices making it a frugal solution to this kind of problems. We present some preliminary results obtained using a real prototype of TurboLoRa.

Wireless links extending beyond the horizon at frequencies of 868 MHz cannot be attributed to ionospheric reflections, since those only happen at much lower frequencies. For very long links the propagation can be attributed to the bending of the radio waves due to anomalies in the atmospheric refractivity index. These anomalies are caused by abnormal variations in temperature and humidity versus elevation that result in tropospheric ducts that can reach thousands of kilometers. Leveraging available data from radiosondes that are periodically launched worldwide, it is possible to determine the refractivity index profile and from this the conditions for the existence of a tropospheric duct. We developed a series of Python tools to analyze such links and applied them to assess the propagation mechanism in three cases, reported by other users, that overcome the earth’s curvature obstruction. These tools can be used to determine the presence of ducting conditions at any place in the vicinity of a radiosonde launching site and results are valid at other frequencies as well. They can also be used by people not versed in Python by using the Jupyter Notebook hosted in Google Colab.

Due to rapid growth in urban population and advances in the automotive industry, the number of vehicles is increasing exponentially, posing the parking challenges. Automated parking systems provide efficient and optimal parking solution so that the drivers can have hassle free and quick parking. One of the demanding requirements is the design of smart parking systems, not only for comfort but also of economic interest. With the advancements in the Internet of Things (IoT), wireless sensors-based parking systems are the promising solutions for the deployment. Optimal positioning of IoT enabled wireless sensor nodes in the parking area is a crucial factor for the efficient parking model with the lower cost. In this paper, we propose a novel multi-objective grey wolf optimization technique for node localization with an objective to minimize a localization error. Two objective functions are considered for distance and geometric topology constraints. The proposed algorithm is compared with other node localization algorithms. Our algorithm outperforms the existing algorithms. The result shows that localization error is reduced up to 17% in comparison with the other algorithms. The proposed algorithm is computationally efficient due to the choice of fast converging parameters.

The growth of the Internet worldwide has been fuelled by the development of the “National Research and Education Networks” (NRENs), i.e., networks of academic and educational institutions. In Africa the establishment of NRENs is more recent. In this paper we analyse the readiness of African NRENs to be part of “The Things Network” (TTN), a network of IoT gateways that has fostered the growth of IoT in Europe by adopting a community network model. We analyse RTT and packet loss toward the nearest TTN network server, in African countries where RIPE Atlas (RIPE - “Réseaux IP Européens”, French for “European IP Networks”) probes are hosted both in academic and commercial networks. Our conclusion is that NRENs and commercial ISPs are on an equal foot in hosting TTN gateways in most countries we considered.

In this paper we present a flexible protocol based on LoRa technology that allows for the transfer of “content” to large distances with very low energy. LoRaCTP provides all the necessary mechanisms to make LoRa reliable, by introducing a lightweight connection set-up and ideally allowing the sending of an as-long-as necessary data message. We designed this protocol as a communication support for edge based IoT solutions given its stability, low power usage and the possibility to cover long distances. We present the evaluation of the protocol with various sizes of data content and various distances to show its performance and reliability.

Traffic congestion is a major problem in many cities. It happens due to the demand-supply imbalance in the transportation network and poor management. Traffic flow slows down when the number of vehicles that travels on the road increases or the roadway capacity decreases due to various reasons. In order to solve this problem, different solutions are proposed to provide reliable, real-time transport management services in an Intelligent Transportation System (ITS). In this paper, we propose a novel real-time video analytics using low-cost IoT devices and LoRaWAN networks to realize new services and applications that include traffic management through IoT edge computing. The use of LoRaWAN for such application is our main contribution. We retrain YOLO v3 object detection machine learning model (transfer learning) for vehicle detection and counting, to make it lightweight and fast enough to be able to run on a Raspberry Pi, a single-board computer with limited RAM. The edge node, with low-cost smart camera and connectivity through LoRaWAN networks counts the number of vehicles using real-time video analytic and report only traffic count to the server. This experimental work provides insight into the applicability of a low-cost IoT system to traffic management with a resource-constrained environment. Real-world video analysis of vehicle detection and counting show the effectiveness of the designed solution. The results demonstrate the effectiveness of the proposed approach.

Connectivity is the essential requirement for a smart city, and traditional techniques are too costly to connect a great number of devices which might be served with lower data throughput, but require longer ranges to accommodate the requirements of IoT. Although cellular technologies and Wi-Fi have proved an enormous success in connecting billions of users to the Internet, they are not the best to connect objects with limited throughput requirements, limited processing power and low power consumption requirements. These needs are being addressed by Low Power Wide Area Networks (LPWAN) that use unlicensed frequencies and by newer cellular protocols aimed at lowering the cost of the devices by limiting their bandwidth capabilities in the licensed bands. This paper discusses these technologies and the choice made to provide an IoT infrastructure for the city of Maputo in Mozambique.

In the IoT world, establishing a strong mobile network architecture will be critical for organizations to bring together people, processes, data and things. Among the various available protocols and standards to network IoT entities, the Message Queue Telemetric Transport (MQTT) is already a reference solution. It provides a publish/subscribe messaging transport specifically designed to be used in devices with limited resources over constrained networks. MQTT's main limitation is its low resilience with respect to device mobility, so that the connections could suffer frequent and long lasting disruptions or high bit error rates that severely degrade normal communications. In this work we propose an architecture to increase the robustness of MQTT by integrating a Disruption Tolerant Network (DTN) approach. The architecture has been evaluated through several experiments using real devices to validate its feasibility, and to derive some guidelines for its use.

Despite successful pilots in developing countries, TVWS (Television White Spaces, unused portions of television broadcasting spectrum) has not gained the amount of attention that it deserves. One of the reasons is the lack of spectrum measurements that would convince the regulators about the abundance of idle spectrum. Spectrum measurements traditionally require expensive instruments and considerable operator's expertise. This has changed with the emergence of low cost spectrum analyzers and smartphones. In this paper we present RFTrack, a software suite that consists of an Android application and a TVWS analysis server. Together, they measure spectrum using a low-cost spectrum analyzer and geotag the data using the phone's internal GPS. The Android application is very easy to use and campaigns can be carried out by non-technical users. Once an Internet connection is available, the program sends data to the server that performs the required processes to present the results in an easy-to-understand way, also allowing for some user customization. The system has been used in eleven countries. We present the results from a measurement campaign in Costa Rica. We believe that this is a useful tool to demonstrate the existence of underutilized spectrum, especially in rural areas of developing countries.

This demo paper presents Cloudy, a Debian-based distribution to build and deploy clouds in community networks. The demonstration covers the following aspects: Installation of Cloudy, the Cloudy GUI for usage and administration by end users, demonstration of Cloudy nodes and services deployed in the Guifi community network.

This paper describes an operational geographically distributed and heterogeneous cloud infrastructure with services and applications deployed in the Guifi community network. The presented cloud is a particular case of a community cloud, developed according to the specific needs and conditions of community networks. We describe the concept of this community cloud, explain our technical choices for building it, and our experience with the deployment of this cloud. We review our solutions and experience on offering the different service models of cloud computing (IaaS, PaaS and SaaS) in community networks. The deployed cloud infrastructure aims to provide stable and attractive cloud services in order to encourage community network user to use, keep and extend it with new services and applications.

TV White Spaces (TVWS) technology and regulation has the potential to make connectivity both technically and economically feasible in rural Africa where affordable access remains a challenge. The superior propagation characteristics of TVWS technology make it particularly well suited to connecting remote communities. Evidence collected from our measurements in Malawi and Zambia suggest that most UHF spectrum is already available both in urban as well as in rural areas and could be used to provide Internet connectivity. In this paper we present the findings of a TVWS spectrum measurement initiative in Malawi and Zambia. We introduce an open hardware device that geo-tags spectrum measurements and saves the results on a micro SD card. The device can also be used to record the use of spectrum over long periods of time.

It is widely recognized that white spaces identification is an important milestone for the wide deployment of next generation cognitive wireless networks. However, spectrum holes detection tools used for white spaces discovery are still either in the infancy stage or too expensive to enable massive white spaces exploitation. Building upon cheap hardware equipment, this paper presents experiments conducted in the town of Trieste in Italy to sense the environment and find out which frequencies are not being used in a particular place and time-of-the-day. As a a step towards white spaces exploitation, we believe that our experimental frequency exploration is an important milestone upon which white spaces patterns recognition will be built with the aim of using these patterns in wireless network planning and management.

The development of a wireless sensor network (WSN) gateway is challenging for sites where limited infrastructures lead to frequent power shortages and and network unreliability. In this paper we presents a low-power, low-cost, 802.15.4 and 802.11 compatible solution which uses open source software to meet local conditions. Using the SunSPOT motes on a system which is mostly platform independent, our system is based on the Fox embedded Linux board and equipped with a USB flash drive and a USB WiFi adapter. The system can be solar-powered, and the results of a solar system design are presented. All the hardware components are available off-the-shelf and are easy to assemble. We conclude that our system is preferred for applications in remote areas, where a stable power supply and a reliable network infrastructure are lacking. Furthermore, it can be used to extend the range of wireless sensor networks by layering a network of long range motes above islands of low range motes.

The last two years have witnessed the introduction and rapid expansion of products based upon large, systematically-gathered, street-level image collections, such as Google Street View, EveryScape, and Mapjack. In the process of gathering images of public spaces, these projects also capture license plates, faces, and other information considered sensitive from a privacy standpoint. In this work, we present a system that addresses the challenge of automatically detecting and blurring faces and license plates for the purpose of privacy protection in Google Street View. Though some in the field would claim face detection is “solved”, we show that state-of-the-art face detectors alone are not sufficient to achieve the recall desired for large-scale privacy protection. In this paper we present a system that combines a standard sliding-window detector tuned for a high recall, low-precision operating point with a fast post-processing stage that is able to remove additional false positives by incorporating domain-specific information not available to the sliding-window detector. Using a completely automatic system, we are able to sufficiently blur more than 89% of faces and 94 - 96% of license plates in evaluation sets sampled from Google Street View imagery.

The Collaborative Sensing Language (CSL) is a high-level feedback control language for mobile sensor networks (MSN). It specifies MSN controllers to accomplish network objectives with a dynamically changing ad-hoc resource pool. Furthermore, CSL is designed to allow the updating of controllers during execution (patching). This enables hierarchical control with simpler controllers at lower levels. The CSL Execution Engine contains the intelligence to allocate resources to tasks dynamically and adjust in real time to resource motion, this enables CSL controllers to be simple, intuitive and scalable. Experimental results show that the CSL Execution Engine performs these services with the addition of very little overhead.

This paper is about the modular compilation and distribution of a sub-class of Simulink programs [9] across networks using bounded FIFO queues. The problem is first addressed mathematically. Then, based on these formal results, a software library for the modular compilation and distribution of Simulink programs is given. The performance of the library is given. The value of synchronous programming for the next generation of traffic control is discussed. The adoption of these tools seems to be the natural candidate to address the needs of traffic engineers. As a case study we present an implementation in Simulink of a controller for coordinated traffic signals in an asymmetric peak hour traffic scenario and we evaluate its computational performance in a distributed environment.