ResearchSpace

Communities in ResearchSpace

Welcome to ResearchSpace, the institutional repository of the CSIR. ResearchSpace is an open access electronic archive collecting, preserving and distributing scholarly digital materials created by the CSIR.

Recent Submissions

Item

Advanced Production & Security

(2026-08) Koen, Hildegarde S

This presentation examines the strategic adoption of Artificial Intelligence (AI) within the Advanced Production and Security (APS) division of the CSIR, with a particular focus on balancing innovation, competitiveness, and the protection of sensitive information. It highlights the growing integration of AI tools into research, engineering, and operational environments, while emphasizing the significant risks associated with cloud-based AI systems, including data leakage, intellectual property loss, regulatory concerns, and the exposure of defence and sovereign information. A three-tier AI adoption model is proposed, comprising Public AI, Enterprise AI, and Sovereign/Secure AI environments, to ensure that AI usage is aligned with information sensitivity and security requirements. The presentation advocates for a “move the AI to the data” approach, whereby sensitive information remains within trusted environments. It further identifies strategic opportunities for the CSIR to become a national leader in Trusted Sovereign AI through the development of air-gapped large language models, secure knowledge management systems, and AI assurance capabilities. The document also outlines short- and medium-term AI value propositions across defence, mining, manufacturing, robotics, and cybersecurity sectors, and presents research themes and practical use cases demonstrating the transformative potential of AI in secure and mission-critical applications. Overall, the presentation underscores the importance of responsible, secure, and human-centred AI adoption to advance research, innovation, and national strategic objectives.

Item

Decentralized elliptic curve identity for authenticated software defined radio mesh messaging

(2026-09) Sebetoa, Matlapane R; Nyareli, Teboho N; Thaba, Janes M

Software Defined Radio mesh networks are useful for disconnected and mobile communication. However, a receiver still needs a way to verify the sender, ensure that the message was not changed, and reject old messages that are replayed. This paper combines a structured literature review with an offline experiment for a decentralized elliptic curve identity layer. The review covers security implementations for the Software Defined Radio physical layer, physical layer authentication and key generation, authentication in mobile and vehicular ad hoc networks, and trust systems based on blockchain. It identifies a gap between radio-layer protection, transmitter authentication, and cryptographic authentication at the message level for Software Defined Radio mesh systems. The proposed layer derives node identities from public keys, signs outgoing payloads, and verifies the trusted key list, signature, and replay state after the radio frame is recovered. The experiment measures payload growth, Low-Density Parity-Check block impact, payload efficiency, signing time, verification time, and security behavior for tampered, replayed, incorrect-key, and unknown-node messages. The results show that signing and verification complete in microseconds, while packet growth is the main cost for small payloads that make inefficient use of message blocks. The results indicate that decentralized elliptic curve authentication is practical for low-rate authenticated Software Defined Radio messaging when compact encoding, key identifiers, and batching are used to reduce airtime overhead.

Item

A review of software-defined machine learning-based congestion controls in wireless sensor networks

(2026-08) Nthai, R; Kobo, Hlabishi I; Maswikaneng, S; Ndlovu, L

The rapid proliferation of the Internet of Things (IoT) has made wireless sensor networks (WSNs) a cornerstone technology for modern applications, from smart cities to industrial automation. However, the inherent constraints of WSNs, such as limited energy, bandwidth, and computational power, make them highly susceptible to network congestion. This issue leads to packet loss, increased latency, reduced throughput, and significant degradation in quality of service (QoS). Although traditional congestion control mechanisms offer some relief, they often lack the adaptability required for the complex and dynamic nature of modern WSNs. This research examines the critical role of emerging technologies in mitigating congestion in WSNs, focusing on the integration of software-defined networking (SDN) and machine learning (ML). This study provides a review of this integration, often termed ML-SDWSN, analysing how centralised, programmable control of SDN can be enhanced by predictive and adaptive capabilities of ML to create more robust congestion control frameworks. It presents a comparative analysis of traditional methods against these novel SDN/MLbased strategies, highlighting performance improvements in throughput, energy efficiency, and network optimisation. Furthermore, this study identifies key limitations and open challenges, particularly the computational overhead of implementing ML algorithms in SDWSN environments for real-time control. By synthesising the current state of the art and outlining future research directions, this review underscores the significant potential of lightweight, hybrid artificial intelligence (AI)-driven solutions to alleviate congestion, thereby enhancing network capacity, ensuring QoS, and improving the overall efficiency and longevity of these networks.

Item

Hybrid-additive manufacturing cost model: A sustainable through-life engineering support for maintenance repair overhaul in the aerospace

(2020) Oyesola, MO; Mpofu, K; Mathe, Ntombizodwa R; Daniyan, IA

The drive for a more flexible Through-life Engineering servicing technology that can enhance both productivity and profitability for the operation of maintenance, repair, overhaul (MRO) in the aerospace implores harnessing the innovation of hybrid-additive manufacturing (HAM). Hybrid manufacturing system seeks to reduce cost by utilizing the tool-less production principle of additive manufacturing combine with the conventional subtractive manufacturing for product finishing or post-processing. HAM has many merits that make it an attractive option in manufacturing of high performance products as peculiar to MRO businesses. Therefore, this paper aims to establish a cost model derived from a “Time Driven Activity Based Costing” approach for the technological integration of Hybrid system to increase economic competitiveness of highly MRO services in the aerospace industry. Hence, a unique cost formulation method was investigated through analytical technique in cost engineering models for the purpose of manufacturing processes.

Item

In vitro antioxidant and anti-inflammatory effects of plantago major, cannabis sativa and ranunculus multifidus on LPS-stimulated raw 264.7 macrophages

(2026-10) Nxumalo, Precious Z; Gom, Buntubonke; Masuku, N; Theron, Anjo; Mokoka, T; Bareetseng, Andries; McGaw, L

Chronic inflammation is a major contributor to cancer progression, driven by persistent pro-inflammatory signalling and oxidative stress. Plant-derived bioactive compounds offer therapeutic potential by modulating these pathways. This study evaluated the antioxidant and anti-inflammatory effects of leaf extracts from Cannabis sativa L., Plantago major L., and Ranunculus multifidus Forssk. in LPS-stimulated RAW 264.7 macrophages. Extracts were tested for cytotoxicity (MTT assay), nitric oxide (NO) inhibition (Griess assay), cytokine modulation (ELISA), antioxidant activity (DPPH and ABTS assays), and 15-lipoxygenase (15-LOX) inhibition. UPLC-MS analysis was used to identify potential phytochemical constituents. All extracts were non-cytotoxic at 5 - 15 µg/mL concentrations. The extract from R. multifidus had the highest NO inhibition (20%), followed by P. major (17%), while C. sativa exhibited no significant effect. The ELISA assay revealed extract-specific cytokine modulation. P. major strongly suppressed IL-6 and IL-1β and enhanced IL-10 secretion, while C. sativa inhibited IL-6, PTGS2 and moderately reduced iNOS. In contrast, R. multifidus showed variable effects and did not inhibit iNOS and IL-1β. However, it induced variable effects on IL-6 inhibition and IL-10 secretion. The antioxidant assays revealed potent radical scavenging: C. sativa achieved 92.5% DPPH scavenging comparable to ascorbic acid, while ABTS scavenging exceeded 90% for P. major and R. multifidus, closely matching Trolox. All extracts inhibited 15-LOX, with C. sativa showing the highest inhibition (96.2%). UPLC-MS identified the presence of cannabigerolic acid (CBGA) and Δ⁹-THC in C. sativa and verbascoside in P. major, while no major compounds were conclusively identified in R. multifidus. These findings provide the first evidence of cytokine modulation and anti-inflammatory activity by R. multifidus. Moreover, they highlight the complementary antioxidant and anti-inflammatory potential of C. sativa and P. major, supported by phytochemical evidence. These extracts represent promising natural immunomodulatory agents.