Browse
Recent Submissions
Item High-performance software-defined radio spectrometer: A GPU implementation for multi-GHz radio astronomical signal processing(2025-12) Gomomo, Lwando; Gaffar, YA; Winberg, SThis paper presents a GPU-accelerated polyphase filterbank (PFB) spectrometer enhanced with spectral kurtosis (SK) capability for real-time operation in radio astronomical signal processing. Modern wideband radio telescopes employ analog-to-digital converters (ADCs) that sample signals at multi GHz rates, generating tens of gigabits per second (Gbps) of data per antenna that demand high-speed processing while operating in radio frequency interference (RFI)-contaminated environments. Existing solutions often separate FPGA- or GPU based spectrometer backends and RFI detection systems, but this work unifies both within a single GPU pipeline. Using NVIDIA’s PyCUDA framework on a Tesla T4 GPU, the system employs a four-tap Hamming-windowed PFB for channelisation, an averager and a higher-order spectral moment estimator based on the SK algorithm for automated RFI detection in the mea sured spectrum. The implementation achieved 8.88 GSamples/s throughput with high spectral fidelity. Tests using radio telescope data confirmed accurate detection of stationary and transient interference while maintaining false-alarm rates only 1.3% above theoretical expectation. This research demonstrates that modern GPUs can enable efficient and flexible high-throughput spectrom eter architectures for advanced digital signal processing (DSP) in radio astronomy.Item AI in the antenna design life cycle: An AI-assisted systematic review(2025-12) Mafa, Ike M; Winberg, SThe increasing integration of Artificial Intelligence (AI) in antenna design presents opportunities for optimizing the design process. This paper presents a comprehensive AI-assisted systematic review of the use of AI in the antenna design life cycle. A Systematic Mapping Study (SMS) is first employed to map the landscape of AI applications in antenna design, identifying key research areas, trends, and gaps. The SMS categorizes studies into three primary domains: AI for Optimization, AI for Synthesis, and AI for Performance Prediction. In the second phase, a Systematic Literature Review (SLR) is conducted to critically evaluate and synthesize empirical findings from the selected studies. The SLR examines the various AI techniques used in each step of the lifecycle. The results from both methods provide a comprehensive overview of AI’s current applications in the field, highlighting key challenges and opportunities for future research. This hybrid approach, combining AI-assisted SMS and SLR, offers both a broad mapping of AI applications and a deep, critical analysis of their real world effectiveness, making it a valuable contribution to the field of AI-enhanced antenna design and AI-assisted systematic reviews.Item Beyond posters: A user-centric digital twin framework for cybersecurity awareness(2026-03) Makharamedzha, Fhatuwani; Baloyi, Errol; Mmbodi, Rendani; Hlongwane, Ndabezinhle ETraditional cybersecurity awareness (CSA) methods, such as posters, flyers, and static training modules often fail to engage users or drive lasting behavioural change. To address these limitations, this paper proposes a novel, user-centric approach to CSA using Digital Twin (DT) technology integrated with machine learning (ML). The proposed framework introduces the concept of a User-Centric Digital Twin (UCDT)-CSA, a dynamic digital replica of each user modelled on their cybersecurity knowledge, behaviours, and risk profile. While UCDTs have been applied in domains such as construction, aquaculture, and healthcare, this work pioneers their use in the cybersecurity context. The system begins with a pre-assessment to capture individual user responses, which are used to configure a personalized training path. Through ongoing interaction with adaptive simulations and scenario-based learning, the UCDT-CSA evolves in real time, enabling training that continuously adjusts to user performance and behaviour. ML models analyse these interactions to refine each twin’s profile, delivering increasingly targeted content and interventions aimed at improving secure behaviours. This approach transforms CSA from a static, compliance-focused exercise into an engaging, data-driven, and behaviourally adaptive learning experience. The paper outlines the architecture of the UCDT-CSA framework, discusses key implementation considerations, and sets the stage for future empirical validation and deployment in government, Small and Medium-Sized Enterprises (SMEs) and academic environment.Item Pipeline for efficient quantization of large language models for resource-constrained deployment(2025-12) Tchiwewe@csir.co.za; Onumanyi, Adeiza JRecent years have seen breakthroughs in large language models (LLMs), such as the GPT and LLaMA family of models that have transformed natural language processing. Despite this, their considerable computational and memory requirements inhibit their deployment in edge and mobile environments. In this paper we introduce a modular quantization pipeline that reduces the memory footprint of LLMs while preserving core performance. We evaluate the basic and advanced quantization techniques, including Absolute Maximum Quantization, Zero-Point Quantization, GPTQ, and NF4, and make use of popular toolkits that include bitsandbytes and AutoGPTQ. Experimental results on representative tasks show that 4- and 8- bit quantized models can be run on commodity GPUs and CPUs with acceptable quality loss. Our experiments demonstrated up to 8x compression, making our pipeline suitable for LLM deployment in both edge and industrial scenarios.Item Enhanced mobile broadband validation over a slice-aware 5G testbed for EdTech applications(2025-12) Mamushiane, Lusani; Makhosa, T; Motsotso, S; Kobo, HThis work demonstrates an education-focused 5G architecture that delivers comparable streaming quality to geographically separated learners over the same network slice. A centralized core with distributed user-plane functions (UPFs) near Limpopo and Cape Town was evaluated using controlled TCP/UDP traffic. Results show stable, site-independent streaming performance at each campus and fairness when both sites transmit simultaneously on the shared slice. TCP experiments with increasing concurrency exhibited predictable capacity sharing without degrading cross-site experience, while UDP baselines reflected low jitter and minimal loss consistent with smooth video playback. These findings indicate that a single logical slice, anchored by regional UPFs, can provide consistent Quality of Experience across provinces, supporting equitable access to digital learning. Practical tooling constraints limited large-scale parallel UDP evaluation, suggesting future validation with multi-device trials and alternative traffic generators.Item AI enabled industrial internet of Things (AI-IIoT) systems: An overview(2026-03) Vishnu, S; Rajagopal, V; Kirubaraj, AA; Sirimella, P; Abu-Mahfouz, Adnan MIThe Industrial Internet of Things (IIoT) plays a key role in transforming the traditional industrial infrastructure into a smart integrated framework through the amalgamation of physical entities and virtual entities. The integration of AI into the IIoT systems opened a new paradigm named AI-IIoT that enhanced the performance of IIoT systems through data centric insights, prognosis, adaptive actuation, and context awareautomation. However, there are many critical challenges that prevents the large scale adoption of the AI-IIoT systems. This paper presents an overview of AI-IIoT systems focusing on its architecture, applications, challenges and future directions.Item Applications of artificial intelligence of Things (AIoT): An overview(2026-03) Vishnu, S; Rajagopal, V; Kirubaraj, AA; Sirimella, P; Abu-Mahfouz, Adnan MI; Ramson, SRJConventional monitoring systems have been revolutionized with the advancements in capabilities of IoT and AI in terms of data collection and decision making respectively. This is coined as AIoT (Artificial Intelligence of Things). This paper conducts study on various applications of AIoT in diverse domains, underlining the developments in technology, challenges, and future possibilities. The focused application areas are industrial automation, crowd monitoring, disaster management, military, agriculture, waste management, health care, smart grid, and transportation. The detailed review is conducted in the paper depicting the role of AIoT in transforming conventional systems to autonomous and intelligent systems.Item Towards an evidence base to support power-to-X (PTX) decision-making in South Africa: Applying systems thinking, knowledge co-production and spatial analysis(2026-06) Snyman-van der Walt, Luanita; Schreiner, Gregory O; Lochner, Paul ADe-fossilisation is a priority, globally and in South Africa. Power-to-X (PtX) technologies could contribute greatly to achieving these ambitions. South Africa’s renewable energy resources, land availability, platinum group metals resources, and port infrastructure, position it as a potential competitor in the global PtX economy. In addition to defossilisation, a domestic PtX economy could make substantial contributions to job creation, improve local livelihoods and facilitate a Just Energy Transition. Vast technologies and infrastructure are required to create the electricity and water inputs to deliver PtX products (for domestic use and export), which, if developed at a sufficient speed, scale, and intensity, could have cumulative, unforeseen consequences. We applied systems thinking, knowledge co-production and spatial analysis to develop a foundational evidence base for future planning, assessment and decision-making on PtX projects towards the sustainable and responsible establishment of a South African PtX economy.Item Enhancing infrastructure maintenance through technology-driven collaboration(2025-07) Mashaba, BN; Mashaba, Hasane P; Sallie, Ismail M; Roux, Michael PCollaborative technologies have transformed information and data sharing within and across organizations. The transformative ease of cross-organization collaboration has amplified the opportunity for inter-governmental collaboration. The South African transport/infrastructure sector can benefit from effective and responsive systems underpinned by collaborative technologies. This paper examines the application of collaborative technology within South Africa’s transport sector, focusing on Gauteng Province’s partnership with the Council for Scientific and Industrial Research (CSIR) to address road maintenance issues. This partnership resulted in mobile applications, enabling citizens to report road issues directly to authorities and receive responses. The study assesses these technologies as examples of targeted innovation addressing practical needs in infrastructure maintenance. The study concluded with findings suggesting that user-centric app solutions effectively facilitate communication between the public and authorities, demonstrating the power of technology-enabled collaboration to enhance public services and infrastructure upkeep.Item Phytoplankton community composition in inland waters from remotely sensed hyperspectral data(2025-05) Sharp, SL; O’Shea, RE; Cortés, A; Forrest, AL; Kravitz, J; Lain, Lisl; Mpaoane, S; Mudzielwana, R; Mudzielwana, R; Mudzielwana, R; Pillay, H; Pindihama, G; Schladow, GS; Smith, Marié E; Torres-Perez, J; Guild, LSPhytoplankton Community Composition (PCC) is an important measure of the aquatic health of inland water bodies. Globally, PCC in inland waters is shifting towards Cyanobacteria dominance, resulting in toxic Harmful Algal Blooms. As such, tools for monitoring PCC are important for management of these water resources. More readily available hyperspectral data from imaging spectrometer missions will allow for PCC identification. This study evaluates the performance of the PCC classification algorithm Phytoplankton Detection with Optics (PHYDOTax) [1] with new application to inland waters in California and South Africa.Item The evolution of penetration testing in the era of AI(2026-03) Baloyi, Errol; Letshwenyo, Mpho; Mtshali, Manello L; Ramantswana, Thanyani AOver the past several decades, penetration testing has transitioned from a predominantly manual, expert-driven activity to a mature discipline supported by automation, modular frameworks, and artificial intelligence (AI)-assisted tools. This study provides a descriptive review of the historical evolution of penetration testing tools, highlighting the major technological and methodological advancements that have shaped the field. In addition, a practical comparative evaluation of two widely used tools, Burp Suite Professional and the Open Worldwide Application Security Project (OWASP) Zed Attack Proxy (ZAP) was conducted using a controlled vulnerable web application, Damn Vulnerable Web Application (DVWA), to assess their performance and usability in a realistic testing environment. The study further examines the impact of AI on the contemporary and emerging landscape of penetration testing tools. The findings suggest that AI is augmenting existing tools through enhanced automation and more effective vulnerability identification, while simultaneously enabling new paradigms in both offensive and defensive cybersecurity practices. This work contributes to the understanding of the evolving role of penetration testing in an AI-influenced context and discusses the implications of these developments for researchers, practitioners, and tool developers.Item Application of engineering management principles to energy systems modelling projects(2025-07) Lomko, Kabelo D; Eboule, PP; Pretorius, J-HThis paper investigates how the integration of engineering management principles can enhance the efficiency, robustness, and impact of energy systems modelling projects. Using South Africa’s evolving electricity sector as a contextual anchor—marked by a legacy dependence on coal and a recent increase in renewable energy deployment—the study addresses persistent challenges in energy systems modelling projects, including data quality limitations, technical complexity, and stakeholder misalignment. The study suggests that these challenges can be mitigated through a structured application of engineering management principles across the energy systems modelling project lifecycle. Employing a mixed-methods approach, the study synthesizes quantitative and qualitative insights from 55 energy sector professionals. The findings reveal widespread support for a formalized framework and highlight the critical role of project planning, risk management, systems thinking, and inclusive stakeholder engagement. Together, these principles offer a cohesive strategy for bridging the gap between technical modelling activities and effective managerial execution in support of sustainable energy planning.Item The effects of platform altitude and terrain type on the false alarm rate in infrared small target detection(2025-10) Malatji, Tsholofelo M; Du Plessis, WP; Bezuidenhout, DF; Nana, Muhammad AThere are many challenges in the field of infrared (IR) small target detection due to the noise-like characteristics of the target. While the development of detection algorithms continues, very little research has sought to understand what creates false targets in IR scenes. The aim of this research is to identify the main factors that contribute to false target generation in the field of IR small target detection. Scenarios with cluttered backgrounds were used to evaluate the effect of a flying platform on the false alarm rate. Both urban and rural scenes were evaluated at different platform altitudes. The study found that IR clutter is generally higher in urban scenes than in rural scenes and that an increase in altitude results in increased false targets. Future work for this study involves the investigation of specific materials in the urban scene that result in the generation of false targets, and scenario conditions such as weather, view angle, time of day, etc. that lead to higher false targets.Item Advancing Innovative Cybersecurity Solutions and Approaches to Protect Digital Ecosystems(2025-12) Mtsweni, Jabu S; Kanyane, M; Phahlamohlaka, Jackie; Munyoka, W; Thomson, K-L; Lynn Futcher, L; Jansen van Vuuren, JIt was our great pleasure to welcome researchers, practitioners, and policymakers to the first IFIP-UNIVEN-CSIR International Conference in Cybersecurity (IFIP-UNIVEN CSIR ICC 2025), held in Pretoria, South Africa, from December 11 to 12, 2025. The conference was jointly organized by the University of Venda (UNIVEN) and the Coun cil for Scientific and Industrial Research (CSIR) under the auspices of the International Federation for Information Processing (IFIP) and focused on the theme: “Advancing innovative cybersecurity solutions and approaches to protect digital ecosystems.” This proceedings volume, published in the esteemed IFIP Advances in Information and Com munication Technology (IFIP-AICT) series, focuses on the technical papers presented at the main conference. The selection process for the research papers was highly rigorous, following quality checks in place. The conference received a total of 43 full-paper submissions. Each paper underwent a thorough double-blind peer-review process with an average of three reviews via the EasyChair system to ensure the highest quality, integrity, and relevance. Based on the reviewers’ recommendations, only 17 papers were accepted for publication and presentation, resulting in an acceptance rate of approximately 39.5%. Following the technical acceptance, all papers were subjected to a final quality check, which included similarity reporting using Turnitin and iThenticate. Six (6) accepted papers required subsequent corrections to address issues identified in the similarity reports, and all issues were addressed to the satisfaction of the editorial team. After the initial reviews and feedback sent by authors, two (2) rejected papers were withdrawn by the authors.Item Reinforcement learning-guided de novo drug design: A comparative study of RL algorithms for small molecule generation(2025) Mpofu, Kelvin T; Thwala, Nomcebo L; Thobakgale, Setumo L; Mthunzi-Kufa, PWe present a comparative study on the application of reinforcement learning (RL) algorithms for de novo drug design. Using a custom molecular environment, we benchmarked five RL methods, DQN, PPO, A2C, REINFORCE, and DoubleDQN, for their ability to generate small, drug-like molecules from atomic building blocks. The models were evaluated based on chemical validity, drug-likeness (QED), molecular complexity, compliance with Lipinski’s Rule of Five, and structural similarity to known pharmaceuticals such as Aspirin and Ibuprofen. Among the tested algorithms, REINFORCE and PPO outperformed others by generating chemically diverse and pharmacologically relevant compounds, achieving the highest QED scores and producing molecules with complex ring structures and higher scaffold novelty. All models successfully generated fully Lipinski compliant molecules, demonstrating their utility in producing viable drug candidates. This work offers insights into the performance dynamics of RL models in chemical space and provides a foundation for developing AI-driven pipelines for accelerated drug discovery. This study highlights the benchmarking gap in RL-based molecule generation and systematically evaluates five algorithms under identical conditions to identify strengths and trade-offs.Item Streamlining Store Separation Analysis with Missile Datcom and Aerodynamic Store Segmentation(2024-07) Ndlovu, Hlamulo P; Jamison, Kevin A; Mthembu, Ndumiso M; Ndebele, Bright B; Zwane, LindokuhleThe focus of the presentation is to describe the process taken in developing a methodology for subsonic flow store separation to quicky generate extensive segmented lookup tables using Missile Datcom to support the fast store trajectory calculation using an inhouse six-degree-of-freedom (6-dof) solver called ARUV.Item Bridging the digital divide in the Republic of South Africa: The emergence of low earth orbit networks(2025-11) Makondo, Ntshuxeko; Kobo, Hlabishi I; Mboweni, Lawrence S; Mathonsi, TEOvercoming the digital divide in rural and remote areas of the Republic of South Africa (RSA) has been a challenging and daunting. This is because of the country's vast geographically landscape. As of 2023, only 70% of South Africans had reliable internet access. The COVID‑19 pandemic has further worsened this gap, as education, business, government services were conducted online. The need for internet has risen significantly as the country is embracing the potential of Information and Communication Technology (ICT) as a stepping stone to economic and social development. However, the traditional way of deploying broadband is limited by the prohibitively expensive nature of extending high‑capacity fibre and microwave backhaul to remote districts, making many business cases unviable for terrestrial operators. As a result, this paper examines the role of Non-Terrestrial Networks (NTNs) specifically Low Earth Orbit (LEO) satellites in bridging this digital divide. Furthermore, this paper examines two promising LEO satellite-based solutions. The first solution leverages LEO constellations as a backhaul for current 5G terrestrial networks. The other solution leverages direct-to-direct (D2D) LEO services to provide low-latency Internet access in remote and underserved areas. This paper further presents the challenges that are slowing down the adoption of LEO, including the regulatory barriers and high deployment costs. The recommendations to expedite LEO adoption and integration into 5G networks are also highlighted. Integrating 5G infrastructure sharing with LEO satellite networks reduces deployment costs, improves rural broadband coverage, and guides policy reforms that promote equitable access and efficient spectrum use in South Africa. This study enhances technical understanding of LEO deployment and provides a strategic reference for policymakers, researchers, and industry leaders working to bridge the rural digital divide.Item Shedding light on loadshedding with natural language processing: A social media case study on public perspectives of the South African electricity crisis in 2022(2025-11) Moodley, Avashlin; Naidoo, PrivolinIn times of collective discomfort and dissat isfaction, people often find solace in shared adversity on social media platforms like X (for merly known as Twitter). These platforms offer a unique window into the public’s emotions and viewpoints concerning common challenges. In 2022, South Africa experienced an electricity crisis, during which the country was subjected to rolling blackouts, commonly known as load shedding, by Eskom, the country’s primary electricity provider, to prevent a national elec tricity grid shutdown. This study conducted a data-driven exploration of the public discourse surrounding Eskom and loadshedding on X us ing natural language processing and data sci ence techniques. The dataset utilised for this study comprised tweets containing keywords related to Eskom and loadshedding. The study delved into the topics of discussion by apply ing topic modelling techniques to uncover la tent themes within the discourse. The topics were analysed through a multifaceted lens to unpack and highlight patterns within the sen timents, emotions and biases that underpin conversations related to loadshedding and Es kom. A notable inclusion in the analysis was the incorporation of sarcasm classifications, which enhanced the interpretation of the emo tion and sentiment within the topics discussed. The findings uncovered from the analysis were contrasted with loadshedding-related events in 2022 to understand the public discourse as the electricity crisis escalated. The methodology of this study provides a framework for utilis ing natural language processing techniques to uncover and examine the perspectives of a col lective within discourse related to events of shared interest.Item Peer-to-Peer based indoor localization using smartphones: A wi-fi RSSI and fingerprinting algorithm approach(2025-12) Sediela, MS; Gadebe, Moses L; Kogeda, OPThe evolution of smartphones with advanced wire less communication network capabilities has accelerated the adoption of Indoor Positioning Systems (IPS). These IPS desire to predict the position or location of various wireless devices. The (GPS) remains the widely adopted Location-Based Services (LB Global Positioning System) application for positioning and navigation in an outdoor setting. However, GPS is inefficient indoors due to the line-of-sight requirements to the satellites. The indoor environment is harsh with multipath effects that cause occlusion between the GPS receiver and transmitter. Short range technologies such as Wi-Fi are gaining popularity indoors to alleviate GPS as an alternative technology. However, Wi-Fi infrastructure can be costly. This paper presents a cost-effective localization solution that utilizes Android smartphones as the sole requirement, eliminating the need for additional hardware. The proposed IPS solution uses a fingerprinting algorithm and employs a Peer-to-Peer (P2P) localization approach to reduce the cost implications of Wi-Fi. Only the received signal strength indicator (RSSI) measurements from Wi-Fi Direct and allied devices are used as input during both the offline and online stages of the fingerprinting process. The proposed IPS developed an Android mobile application in Java programming using Android Studio, with SQLite and Firebase real-time Database for storage. We have tested the system in real-time and evaluated its performance; the system produced a high accuracy of 93.33% for monitoring.Item Dynamic deformation behavior of the TM380 mild steel subjected to blast(2025-06) Shoke, Lerato S; Sono, Tleyane J; Mutombo, Kalenda; Snyman, IMThe dynamic deformation behavior of the mild-steel TM380 subjected to explosive loading has been investigated. An imparted impulse and high pressure, from a PE4 explosive charge, interacted with the plate which is attached to a deflection gauge designed to measure the mid-point deflection time history and the imparted impulse. The shape of the bulge at the midsection of the plate was that of a paraboloid. The deflection-time curve is characterized by an escalation, followed by a very short plateau of a few microseconds at mid-point deflection, and finally a drop in deflection timespan. The dynamic strain, strain rate and impulse changes are revealed by deflection-time, velocity-time and hydrostatic pressure curves. Although no significant change in grain size and morphology occurs after shock wave loading, the pearlite lamellar structure transformed into spheroidized cementite as a result of shock induced phase transformation.