The use of Television White Spaces (TVWS) is globally gaining traction. TVWS technology was in its infancy a decade ago. Many authorised trials conducted by several countries around that time were all successful. Kenya conducted its first TVWS trial in 2013. The trial demonstrated that TVWS could successfully deliver broadband access with no interference to Digital Terrestrial Television (DTV). Mawingu, which was the pilot project on TVWS in Kenya, also exemplified that the technology can help bridge the digital divide. To drive commercial deployments for TVWS, Communications Authority of Kenya (the Kenyan Regulator), has recently drafted regulations on the use of TVWS. Hence, this paper presents an overview of the regulatory framework for Kenya on TVWS. It also compares and analyses the Kenyan TVWS framework with other existing frameworks in Africa. Additionally, it also extends a comparison with the regulatory models of the United States, the United Kingdom and a select set of other countries. The paper further gives recommendations on the adoption and implementation of TVWS regulations for national regulatory authorities (NRAs) in Africa.
Authors: Kennedy Ronoh (Technical University of Kenya), Leonard Mabele (Strathmore University), Dennis Sonoiya (University of Strathclyde),
Hide Authors & Abstract

Due to interoperability requirements among the heterogenous networks (Hetnets), the security requirements for 5G are high compared to 2G, 3G and 4G. The current 5G handover protocols are based on either fuzzy logic (FL), artificial neuro networks (ANN), blockchain, software define network (SDN), or Multi-layer Feed Forward Network (MFNN). These protocols have either long latencies or focus on either security or quality of services parameters such as user satisfaction. The usage of these inefficient authentication schemes during 5G handovers lead to performance degradation in heterogeneous cells and increases the delay. Apart from increased handover delays, 5G networks experience frequent failures of the handoff process, both of which reduce capacity gains offered by 5G networks. Consequently, the provision of strong security, privacy and low latency handovers is required for the successful deployment of 5G networks such as wireless local area networks (5G-WLAN) heterogeneous networks. The goal of this paper was therefore to develop an ANN-FL protocol that addressed both security and QoS 5G network issues. The simulation results showed that the developed protocol was robust against attacks such as eavesdropping, de-synchronization and tracing attacks and yielded a 27.1% increase in handover success rate, a 27.3% reduction in handover failure rate, and a 24.1% reduction in ping pong handovers.
Authors: Vincent Nyangaresi (Tom Mboya University College), Anthony Rodrigues (Jaramogi Oginga Odinga University of Science & Technology), Silvance Abeka (Jaramogi Oginga Odinga University of Science & Technology),
Hide Authors & Abstract

The Digital Video Broadcasting-Terrestrial, second generation (DVB-T2) system is now mature and being deployed worldwide in replacement of the first generation (DVB-T) or in direct deployment. Nevertheless, attempts to improve its performance in terms of distance to Shannon limit, Bit Error Rate (BER), Signal to Noise Ratio (SNR) or coverage are still reported in the literature. On the one hand, the authors of this paper reported recently that Universal Filtered Multicarrier (UFMC) is, among 5G waveforms, the best compromise in terms of improvement, spectrum efficiency and complexity for the replacement of Cyclic Prefix - Orthogonal Frequency Division Multiplexing (CP-OFDM) in DVB-T2 system. On the other hand, a gain in DVB-T2 performance in Additive White Gaussian Noise and Rayleigh environments was reported in the literature using optimized Non Uniform Constellations (2D-NUCs). This paper focuses first on the maximum obtainable performance improvement of DVB-T2 CP-OFDM with NUCs in TU6 environment. It concentrates afterwards on the ultimate gain achievable using joint UFMC and NUCs in DVB-T2. In these conditions, a gain of 0.5 dB (for BER = 3.10^3) is reported in TU6 using CP-OFDM NUC 32K 256-QAM and CR= 1/2 and 3/5 in place of sole CP-OFDM. Also, using in conjunction both technologies, namely UFMC NUC 32K 256-QAM CR = 1/2 and 3/5, a gain of 1.2 dB (for BER = 3.10^3) is achievable which provides a good SNR margin e.g. to increase the emitter’s coverage.
Authors: Anne-Carole HONFOGA (University of Mons, Belgium), Michel DOSSOU (University of Abomey-Calavi (BENIN)), Péniel DASSI (University of Abomey-Calavi, Benin), Véronique Moeyart (University of Mons, Belgium),
Hide Authors & Abstract

The design of a wireless mesh network is usually posed as a multi-objective optimization problem. In this paper, we consider the planning of a wireless mesh network in a rural region where the network coverage and the cost of the architecture must be optimized. In addition, mesh routers are heterogeneous, meaning that they may have different transmission ranges. In the network model, we assume that the region to serve is divided into a set of small zones of various types, including cost-effective locations and zones of interest for which the coverage is mandatory. The objective is then to minimize the number of routers, their types and locations which maximize the coverage percentage of mandatory zones in terms of coverage while minimizing the overall cost of the architecture. To achieve this, we propose three multi-objective approaches. We test the proposed approaches on several random topologies. The min-max regret metrics is used to appreciate the quality of solutions of the Pareto front of different approaches.
Authors: Jean Louis Fendji Kedieng Ebongue (University of Ngaoundere), Christopher Thron (Texas A&M University Central Texas, Killeen, USA), Anna Förster (University of Bremen, Bremen, Germany),
Hide Authors & Abstract

Nowadays Location-Based Services (LBS) applications are proposed and presented in literature because of high growth in wireless sensor networks. The LBS provide useful information about the person or object’s current position. Among these applications, GPS-based positioning have been deployed in an outdoor environment. The GPS fails in an indoor environment because of multipath effects caused by walls or indoor setup in general. Meanwhile, Wi-Fi-based positioning system are being proposed in literature and most utilizes the fingerprinting algorithm. Whereby, measurements of Received Signal Strength Indicator (RSSI) are collected during the offline phase for radio map and positioning is performed during the online phase. Similarly, this approach is faced with signal attenuation challenge caused by walls, desks, moving people, or just the indoor setup in general. The collection of the signal strengths without correction can affect positioning accuracy. In this paper, we present a model that utilizes Mean and Standard Deviation to identify and correct unstable low RSSI outliers during offline phase of Fingerprinting algorithm. We conducted comparative performance of Machine learning classification algorithms based on corrected RSSI and measured RSSI. The results of our simulations in MATLAB, indicates that correcting the RSSI improves the accuracy, precision, and recall of Fine Tree, Coarse Decision Tree, and Kernel Naïve Bayes to 95.1%, 94.8 and 94.5% respectively.
Authors: Madikana Sediela (Lecturer (ICT) and Student in Computer Science), Moses Gadebe (Lecturer in Computer Science), Okuthe Kogeda (Associate Professor: Computer Science and Informatics),
Hide Authors & Abstract

Wireless sensor networks (WSN) consist of a collection of independent sensor nodes that monitor certain conditions in different locations and transfer data through a network to an endpoint called base station. Because WSN nodes are often deployed outdoors where a direct power outlet may not be available, energy efficiency becomes a key factor in the design of WSNs. This work proposes an antenna transmission power optimisation algorithm (TPOA) which dynamically adjusts the power level of the transmission according to the last received signal strength indicator (RSSI). The mechanism was implemented and tested on an Xbee WSN that uses a modified Mixed Hop and signal Received routing for mobile Wireless Sensor Networks (MHRWSN) protocol. The energy consumption results have shown that the proposed system can save as much as 10% of power in Tx operations while maintaining a packet reception rate of above 85%.
Authors: Visham Ramsurrun (Middlesex University Mauritius),
Hide Authors & Abstract

Abstract. The 5G mobile telephony has become one of the worldwide most anticipated technology, which is followed by strong controversy regarding potentially dangerous health effects. This technology relies on electromagnetic field (EMF) emission from its network base stations, increasing the level of existing EMF in the environment. Consequently, this fact has initiated deep concerns of the public, who demanded overall investigation and monitoring of the inevitable 5G EMF exposure. In the last decade, the wireless sensors networks for EMF monitoring emerged as an innovative approach for effective analysis of EMF in the environment. The latest one is the Serbian EMF RATEL network, which offers a sophisticated approach of telecommunication service-based EMF monitoring. This network performs wideband monitoring, counting the EMF contribution of all active EMF sources in a predetermined frequency sub-band. In this paper, the preliminary EMF monitoring of 5G is presented, explaining technical details on the used Narda AMS 8061 sensor, the acquisition process, as well as the analysis and dissemination of the measurement results. The EMF RATEL is envisioned to be a support for the control and management of EMFs in upcoming smart-city ecosystems, in which is expected that will display intensive EMF radiation in living surrounding, regarding various telecommunication services.
Authors: Nikola Djuric (Faculty of Technical Sciences, University of Novi Sad, Trg D. Obradovica, Novi Sad, Serbia), Nikola Kavecan (Falcon-Tech, IT Consulting and Development, Dusana Danilovica 1, Novi Sad, Serbia), Nenad Radosavljevic (Regulatory Agency for Electronic Communications and Postal Services (RATEL), Palmoticeva 2, Belgrade, Serbia), Snezana Djuric (Institute BioSens, University of Novi Sad, Dr Zorana Djindjica 1, Novi Sad, Serbia),
Hide Authors & Abstract

Software Defined Networking (SDN) is no longer a jargon in networking as it is already existent in major data center architectures today. SDN definitely brings along benefits such as manageability and automation of network and management, however , at the expense of major drawbacks such as huge investment in SDN-capable hardware, vendor lock-in and backward incompatibility with legacy devices. SDN itself being a new concept, provides very few aspects in common with traditional networking with each SDN vendor usually limiting the SDN features to their own devices. Even with the introduction of open protocols as OpenFlow aiming to provide vendor neutrality, backward compatibility still remains a problem along with cost, which is why the adoption of SDN has been relatively slow and reluctant. This paper is geared towards addressing the main issues governing the migration towards SDN and hence provide the desired vendor neutrality, backward compatibility without compromising on networking features, security, ease of deployment and management. With this in mind, an SDN Middleware System has been conceptualized to offer the aforementioned features whereby the backend of the system would be responsible to intercept, inspect and process OpenFlow configurations from the SDN Manager/Controller and interpret these commands converting them into the desired configuration in legacy networking terms after which, the legacy nodes are configured with the equivalent of the legacy vendor OS.
Authors: Bhargava Sokappadu (University of Mauritius), Avinash Mungur (University of Mauritius),
Hide Authors & Abstract

Software-Defined Networking (SDN) has become a popular paradigm for modern day optimal performance of the network system as a result of the separation of the control component from other network elements. This enables the maintenance of the flow table structure on these devices while optimal for-warding of packets is enhanced via the central controller. Being a growing net-work architecture which is supposed to be able to meet up with increasing traffic demands in the future, it becomes apparently important that the mechanism that takes care of the QoS of the network demands is put in place. Such demands include the smooth running of big data transmission, D2D video exchange, Voice over IP and real-time multimedia applications which needed certain QoS require-ments for optimal service delivery. However, fewer research articles have re-ported on the improvement on the QoS routing especially in connection with the SDN paradigm. We propose a multi-criteria routing algorithm that is based on deterministic Adaptive rendering technique called DART_MCP. Our DART_MCP QoS routing algorithm deployed Dijkstra’s algorithm to simplify the topology of the network before using multiple-criteria energy function to ad-dress the QoS requirements. We recorded a relatively stable bandwidth and user experience maximization under a low rate of network convergence in comparison with other approaches.
Authors: Ayotuyi Akinola (University of Zululand, KwaDlangezwa, RSA), Matthew Adigun (Centre of Excellence), Pragasen Mudali (Centre of Excellence),
Hide Authors & Abstract

IoT infrastructure makes great demands on network control methods for an efficient management of massive amounts of nodes and data. This network requires fine traffic control management to ensure an adequate QoS for data transmission process, especially in a low-cost network that covers smart territories deployed in so-called "technological lag" areas. Software-Defined Networking (SDN) enables to handle dynamically network traffic as well as flexible traffic control on real-time. However, SDN technology exhibits several issues with regard to additional processing time or loss that are associated to control plan. These factors can lead to performance degradation of the SDN control traffic flows within data plane which is not tolerated in medium/low capacity IoT environment. This paper proposes an Enhanced Flow-based QoS Management approach, called EFQM, that reduces spent time within control plane as well as uses SDN controller either to reduce loss or to optimize bandwidth according to flows latency and bandwidth requirement. Our experimental results show that EFQM outperforms AQRA in terms of response time and packet loss rate. Furthermore, by considering a default routing and delay as metrics, EFQM improves the average end-to-end flow performance by 7.92% compared to AQRA. In addition, EFQM enhances end-to-end flow performance by 21.23% and 23.52% compared to AQRA respectively according to delay and packet loss rate. The measured EFQM runtime is 23.29% shorter than AQRA.
Authors: Avewe Bassene (Université Cheikh Anta Diop), Bamba Gueye (Université Cheikh Anta Diop),
Hide Authors & Abstract

Small-scale farming has an important role in agriculture. Driven by the popularization of the Internet of things (IoT), this paper presents an IoT sensing box prototype coupled with soil analysis through computer vision to help small-scale farmer improve their yields. The idea of combining image-based soil classification with regular soil sensors is to improve the reliability of the sensed data, by minimizing the occurrence of incorrect estimates for specific soil types. The prototype follows a Do-It-Yourself (DIY) approach, and is based on commercial off-the-shelf (COTS) hardware and open source software. Additionally, the prototype includes a casing to house all the hardware which was designed considering standard 3D printing to be easily replicated. As the presented solution is currently on a prototype stage, the validations carried out in a controlled, in-lab environment include guaranteeing the proper process for image acquisition and data quality of the computer vision component (images collected from sensor camera, and their quality), suitable data exchange over the sockets, and the sensing box’s ability to gather data.
Authors: Antonio Oliveira-Jr (Fraunhofer Portugal AICOS & Institute of Informatics (INF) - Federal University of Goias (UFG)), Carlos Resende (Fraunhofer Portugal AICOS), Andre Pereira (Fraunhofer Portugal AICOS), Pedro Madureira (Fraunhofer Portugal AICOS), João Gonçalves (Fraunhofer Portugal AICOS), Ruben Moutinho (Fraunhofer Portugal AICOS), Filipe Soares (Fraunhofer Portugal AICOS), Waldir Moreira (Fraunhofer Portugal AICOS),
Hide Authors & Abstract

With the rapid growth of the Internet of Things (IoT), the deployment, management, and identifi-cation of IoT devices that are connected to networks become a big concern. Consequently, they emerge as a prominent challenge either for mobile network operators who try to offer cost-effective services tailored to the IoT market, or for network administrators who aim to identify as well reduce costs processing and optimize traffic management of connected environments. In order to achieve high accuracy in terms of reliability, loss and response time, new devices discov-ery techniques based on traffic characteristics are mandatory in favor of the identification of IoT connected devices at run-time. Therefore, we design GBC_IoT, a group-based machine learning approach that enables to identify connected IoT devices through network traffic analysis. CBC_IoT framework uniquely identifies IoT devices with an overall accuracy near to 99.98% by considering well-known machine learning tools. Therefore, GBC_IoT can efficiently identify IoT devices with less overhead compared to previous studies
Authors: Avewe Bassene (Université Cheikh Anta Diop), Bamba Gueye (University Cheikh Anta Diop),
Hide Authors & Abstract

LoRa Networks is one of the fast growing and promising techniques to enable communications for Internet of Things (IoT) devices in a large scale or in a long-range communication. In the existing LoRa technology wide range of communication settings is provided, whereas, many combination settings are orthogonal and can be decoded at the gateway in the process of transmitting signals simultaneously. Spreading Factors (SF) plays a major role in enabling long-range multiple reception of packets with every packets being assigned a different spreading factor. Therefore, a change in SF is necessary for improving the data rate for transmission where the link is better and allow LoRa networks to adapt the range trade-off. This work use an open-source framework FloRa for carrying out end-to-end LoRa simulations network in OMNET++ simulator and uses INET framework. In this paper we investigated Adaptive Data Rate (ADR) and provided the behaviour of SF and data rate in LoRa wide Are network (LoRaWAN) as the main findings.
Authors: Smangaliso Mnguni, Pragasen Mudali (University of Zululand), Adnan Abu-Mahfouz (CSIR), Mathew Adigun (University of Zululand),
Hide Authors & Abstract

Remote learning is a system of education, wherein teachers and learners are not in the same location but separated by time and space. Global pandemics such as the Covid-19, necessitate social distance, thereby rendering traditional "contact" based classroom learning unfeasible. e-Learning which is a viable alternative often depends on reliable Internet. Unfortunately, Internet penetration in many areas of the world is still abysmally low. Remote learning in these under-served regions of the world is thus a major challenge. In this paper, we review the state of e-Learning and consider feasible options for under-served communities. Requirement for remote learning, including data and associated cost of attending classes online are also considered. Finally, recommendations for achieving least cost remote learning for under-served communities are given.
Authors: Olasupo Ajayi (University of the Western Cape, South Africa), Hloni Maluleke (University of the Western Cape), Antoine Bagula (University of the Western Cape),
Hide Authors & Abstract

The real estate market in many African countries reflects inefficiency, indiscipline, suspicion, and fraudulent activity. Beyond the direct personal and financial costs, the friction of the existing property transfer process prevents assets from being utilized and valued at their maximum utility. Remarkably, the existing Ethereum blockchain platform and smart contract capabilities can be used to bring efficiency, accuracy, trust, and value to the property transfer and registration process without investment in new infrastructure. We develop a blockchain smart contract that provides transparency and traceability of digital assets from initial registration through all subsequent transfers. While blockchain-based land registration systems have been proposed in other contexts, they are primarily for land title and deed registration. The system modeled in this paper performs these functions while also allowing the owners to trade in the tokenized assets. While we focus on the technical contribution, we also consider the infrastructure, legal, financial, and cultural factors relevant to designing a land registry solution for a developing country. While we focus on the case of Ghana, our solution is applicable anywhere and is particularly appropriate for Africa because it does not require significant new infrastructure. A similar approach may be taken for a variety of e-government services such as digital identity management and online certificate management.
Authors: Isaac Coffie (Carnegie Mellon University Africa), Martin Saint (Carnegie Mellon University),
Hide Authors & Abstract

Big data is large volume of data produced on a daily basis. Distributed storage systems (DSS) is environment that handles, manages and stores those data. The main drawbacks are the lack of system storage capacity, the network device failures that can appear anytime, on time data processing and the system efficiency. All above mentioned issues can be overcome by applying different coding techniques for data distribution. Till this moment many coding schemes are proposed by the researchers. Determining the most efficient code for usage is still a tricky question, which yields an adequate comparison strategy for code selection. The basic Dimakis comparison method offers analysis between the codes regarding the parametes storage per node and download bandwidth to be repair one node. Total comparison method includes in the analysis the total number of nodes in the system together with the overall storage and total downloaded bandwidth in the repair process, with notation that the file size for all codes must be same. In this paper, we are proposing new method for comparison, called Normalized, that enables consideration of broader specter of parameters and not necessarily the same file size of the proposed codes.
Authors: Natasa Paunkoska (University of Information Science and Technology), Ninoslav Marina (University of Information Science and Technology (UIST) “St. Paul the Apostle”), Weiler Finamore (Federal University of Juiz de Fora (UFJF)),
Hide Authors & Abstract

There is relentless effort in combating the issue of crime in South Africa and many parts of the world. This challenge is heightened in under-resourced settings, where there is limited knowledge support, thus resulting in increasing negative perceptions of public safety. This work presents a predictive policing model as an addition to a burglar alarm system deployed in a community policing project to improve crime prevention performance. The proposed model uses feature-oriented data fusion method based on a deep learning crime prediction mechanism. Feed-forward Neural Network (FFNN) and Recurrent Neural Network(RNN) models are employed to predict the amount of calls made to police stations on a monthly basis. Device installation and census data are used in the feature selection process. Coefficient of correlation function is used to isolate the relevant features for the analysis. To provide a viable way of achieving crime reduction targets, the models are implemented and tested on a real-life community policing network system called MeMeZa, which is currently deployed in low-income areas of South Africa. Furthermore, the model is evaluated using co-efficient of determination (R-squared) function and the accuracy of the predictions assessed using an independent dataset that was not used in the models’ development. The proposed solution falls under the Machine Learning and AI applications in networks paradigm, and promises to assist policing in under-resourced settings.
Authors: Omowunmi Isafiade (University of the Western Cape), Brian Ndingindwayo (University of the Western Cape), Antoine Bagula (University of the Western Cape),
Hide Authors & Abstract

Internet security and Quality of Experience (QoE) are two antagonistic concepts that the research community has been attempting to reconcile. To address the prevalence of Cybercrimes, the Internet community has been developing more security features, often at the expense of QoE. One example is the introduction of encrypted Domain Name System (DNS) protocols. These protocols, combined with sub-optimal routing paths and offshore hosting, have the potential to negatively impact the quality of web browsing experience for users in Africa. This is particularly the case in constrained edge networks that are far away from essential infrastructure such as DNS, webservers and security servers. In this paper, we analyse the performance impact of using public DoH/DoT resolvers when resolving websites that are hosted in Africa versus offshore hosting. The study shows that high latency and circuitous DNS resolution paths amplify the performance impact of secure DNS protocols on DNS resolution time and pageload time. The study further shows that users' DNS resolver preferences hugely determine the level of QoE. This study proposes wider adoption of Transport Layer Security version 1.3 (TLSv1.3) to laverage its latenct-reduction features such as false start and Zero Round Trip Time (0-RTT). The study further proposes localisation of content and secure DNS infrastructure. This, coupled with peering and cache sharing recommended by other works, will further minimise the impact DoH/DoH QoE.
Authors: Enock Mbewe (University of Cape Town), Josiah Chavula (University of Cape Town),
Hide Authors & Abstract

Leveraging multiple datasets, we evaluate the current status of African country code Top Level Domain (ccTLD)s, technical environment with regard to best practices. We find that all African ccTLDs meet the minimum requirement of having a minimum of two nameservers (NS) to serve their zone with more than 53% using 7 or more NS. However, some African ccTLDs lack basic DNS configuration like consistent serial numbers or supporting extension mechanism for DNS (EDNS). With regards to the early stage of IPv6 deployment in AFRINIC region, it is interesting to note that 94% of African ccTLDs have nameservers with IPv6 support. This is thanks to out of region or offshore DNS anycast provider which sometimes help to enable DNSSEC. Moreover, African ccTLDs use resources from other RIRs than AFRINIC. Therefore, most African ccTLDs infrastructure are not directly usable by African Internet users.
Authors: Alfred Arouna (Simula Research Laboratory), Amreesh Phokeer (AFRINIC), Ahmed Elmokashfi (Simula Research Lab),
Hide Authors & Abstract

The Domain Name System (DNS) is one of the most critical services for daily operation of the Internet. It resolves names to IP addresses and vice versa on the Internet through a distributed hierarchical system. This work aims at characterizing the authoritative DNS nameservers of two categories of domain (1) publicly available in-addr.arpa and ip6.arpa reverse zones managed by AFRINIC, (2) 57 ccTLDs in the the African region. We study several aspects such as the number of nameservers, the geographical and topological distribution, EDNS and TCP compliance. Overall, the authoritative servers of reverse zones of IP addresses allocated by AFRINIC to their members are 75% EDNS compliant and 72% TCP compliant while the authoritative servers of Africa ccTLDs are respectively 46% and 43.6% compliant for EDNS and TCP. The study also revealed other important information such as the clear domination of some authoritative nameservers, which represents a potential risk of service disruption should these servers become unavailable. Similarly, the geographic location of the authoritative nameservers may potentially have an impact on response times to DNS resolutions and affect user experience. Therefore, a series of efforts must be done in those areas to ensure the optimal functioning of the Internet in the region.
Authors: Yazid Akanho (IGBANET), Malick Alassane (World Internet Labs), Mike Houngbadji (IGBANET), Amreesh Phokeer (AFRINIC),
Hide Authors & Abstract