DSpace Collection:

A hybrid intrusion detection model for application Layer DDOS Attacks based on K-Means and Cart Algorithms

2026-06-23T08:24:38Z

Title: A hybrid intrusion detection model for application Layer DDOS Attacks based on K-Means and Cart Algorithms Authors: Cheruiyot, Victor Kipngetich Abstract: The increase in interconnectivity and advancement in network technologies have influenced a parallel rise in Distributed Denial of Service (DDoS) attacks, and the perpetrators have become sophisticated such that previously dependable tools and techniques have become ineffective. The purpose of the study was to design an intrusion detection model based on K-Means and CART algorithms, and train and test it using the CICDDoS2019 dataset, which represents application-layer DDOS attacks. The objectives of the study were to: Determine the existing application-layer intrusion detection techniques and models; Explore the weaknesses of existing intrusion detection models; Classify the dataset using individual K-Means and CART algorithms; Develop a hybrid intrusion detection model for application-layer DDoS attacks by combining K Means and CART algorithms; and evaluate the performance of the hybrid model. The study was designed as a quantitative experimental simulation. It adopted the empirical positivist paradigm. A machine learning theory and network security theory formed the theoretical framework. The Scikit-Learn libraries were employed using Python programming to perform the analysis. The study utilised secondary data obtained from the CICDDoS2019 dataset, containing 49.59 million records of 12 unlabelled DDoS attack types including NTP, DNS, LDAP, MSSQL, NetBIOS, SNMP, SSDP, UDP, UDP-Lag, WebDDoS, SYN, and TFTP. This research used simple random sampling to select 30000 records from each attack type, yielding a dataframe of 110,000 rows and 88 columns. The Unsupervised component of the experiment requires no training and testing sets. For the supervised component using the CART algorithm, the dataset was split into 67% for training and 33% for testing. Individually, the K-Means algorithm achieved homogeneity, completeness, and V-measure scores of 50.76%, 51.95%, and 51.35% respectively. On the other hand, CART was measured on accuracy, precision, recall/sensitivity, and F1-Score and it achieved scores of 74% on all counts. The hybrid model was fundamentally a CART algorithm improved by K-means clustered features and therefore was scored on the CART algorithm metrics basis. It scored 78% on accuracy, 79% on precision, 78% on recall, and 78.5% on F1-score. The dataset proved to have high dimensionality and complexity with multiple overlapping clusters. K-Means had an average performance proving its unsuitability for this type of dataset. CART algorithm had a relatively high success in identifying application layer DDoS attacks. The hybrid model achieved a better performance score compared to its constituent models as shown by the difference between the chosen metrics and their averages. This study concludes that our hybrid intrusion detection model can outperform existing K Mean and CART algorithms in terms of accuracy, precision, recall and F1 score. The study recommends that future studies should investigate a similar model using density based clustering algorithms like DBSCAN in place of K-Means in a similar setup.

Leveraging Records Management in ensuring access to public information for sustainable Development in Uasin Gishu County, Kenya

2026-06-23T08:05:43Z

Title: Leveraging Records Management in ensuring access to public information for sustainable Development in Uasin Gishu County, Kenya Authors: Kimitei, Mark Kipchumba Abstract: Access to public information is crucial for sustainable development. However, in many countries of the world, especially in developing economies, this is an ideal whose achievement has faced challenges marked by limited technological infrastructure, fragmented records management systems and insufficient resources. In the Kenyan context, studies have revealed that challenges in records management affect access to public information for sustainable development. It was against this backdrop, that the present study sought to assess how records management can be leveraged to ensure access to public information for sustainable development in Uasin Gishu (UG) County. Consequently, the study addressed the following objectives namely to: examine the current records management practices in UG County; establish the link between UG County Integrated Development Plan (CIDP) and access to public information; evaluate the effectiveness of records management in ensuring access to public information as a prerequisite for the attainment of sustainable development in the County; and propose records management strategies that will enhance access to public information held by the County government as a means of promoting sustainable development. The study was anchored on two pivotal theoretical frameworks: The Records Continuum Model and the Process Model of Information Management. It was grounded on pragmatic research paradigm associated with mixed methods approach and intrinsic case study design. The population of the study was 110 respondents comprising, 10 County Executive Committee members (CECs), 15 Chief Officers (COs), and 30 Members of the County Assembly (MCAs) representing the public all of whom drive the development agenda of the County and 55 Records and Clerical Officers (RCOs) charged with the responsibility of records management. Given the small size of the population, a complete enumeration of the population (census sampling) was adopted. Quantitative data was collected from COs, RCOs and MCAs using questionnaires while qualitative data was collected through in-depth interviews with CECs, supplemented by observation and documentary review. Qualitative data was analyzed thematically and presented in a narrative description while quantitative data was analyzed using descriptive statistics. The findings of the study revealed that UG County’s records management practices remain largely paper-based. The study also found that the County’s CIDP is aligned with access to public information for sustainable development, as evidenced by strong institutional commitment to transparency, stakeholder engagement, records management integration, and Sustainable Development Goal (SDG) 16 albeit with implementation gaps. Similarly, the study also found that records management in the County is largely effective in supporting access to public information and sustainable development, though capacity, policy, and digitization gaps constrain full realization of its benefits. It further found that UG County is pursuing ICT-driven records management strategies to enhance access to public information. The study thus concluded that records management plays a critical role in providing access to public information for sustainable development. The study recommends that UG County government develops and implements a comprehensive records management policy, allocate adequate resources for the records management function, and conduct regular staff training in records management.

Influence of communication strategies on promotion of maternal health services in Baringo county, Kenya

2026-06-12T07:00:37Z

Title: Influence of communication strategies on promotion of maternal health services in Baringo county, Kenya Authors: Tuwei, Lilian Cherobon Abstract: Maternal mortality remains a significant public health concern in Baringo County. The maternal mortality ratio (MMR) in Baringo County in 2024 was reported as 488 deaths per 100,000 live births, which was a figure higher than the national average of 374 deaths per 100,000 live births. This was despite the Linda Mama programme being offered freely to improve healthcare access and reduce maternal mortality. Therefore, the study was conducted to investigate the influence of communication approaches on promotion of maternal health services in Baringo County, Kenya in order to recommendations on safe maternal health care. The following objectives guided the study: to assess the influence of mass media on the promotion of maternal health services, to determine the influence of audiovisual media on the promotion of maternal health services, and to establish the influence of interpersonal communication channels on the promotion of maternal health services. The magic bullet theory and the cultivation theory guided the study. The study adopted convergent mixed research methods and descriptive cross-sectional research design. The target population for this study was 6,154 women and 26 health workers. Cluster sampling was used to select a sample of 392 participants, and a census approach was employed in which all 36 health workers were involved in the study. Questionnaires for women and in-depth interviews for health workers were used to collect data. Quantitative data was analyzed using both descriptive and inferential statistics. Qualitative data was analyzed using thematic analysis. The study results revealed that there was a positive linear effect of mass media (β 1 =.167, p<0.05), traditional media (β 2 =.231, p<0.05), audiovisual media (β 3 =.250, p<0.05) and interpersonal communication channels (β 4 =.306, p<0.05) on promotion of maternal health services. The findings from qualitative data revealed that mass media coverage through local radios, community engagement initiatives through chief Barazas and community health workers and customization of the charts, pamphlets and brochures into Kiswahili and Kitugen have improved awareness and attendance of mothers to maternal health services. The study concluded that various forms of media, including mass media, audiovisual media, and interpersonal communication, significantly promoted maternal health in Baringo County. The study recommended strengthening community-based communication approaches through Community Health Volunteers (CHVs) to bridge the communication gap between health facilities and local populations. The study also recommended mobile health (mHealth) solutions, such as SMS reminders and telehealth consultations, to reach women in remote areas.

Integration of artificial intelligence technologies in news production and distribution: a multiple case study of two mainstream media houses in Kenya

2026-02-06T08:48:52Z

Title: Integration of artificial intelligence technologies in news production and distribution: a multiple case study of two mainstream media houses in Kenya Authors: Kandie, Mercy J Abstract: Artificial Intelligence (AI) refers to algorithm-based computational systems capable of mimicking human cognitive functions, such as problem-solving, decision-making, language understanding, and pattern recognition. Its growing use in global media industries has enhanced newsroom efficiency by automating routine tasks, enabling real-time data processing, and supporting personalized content delivery. Despite these opportunities, AI also raises concerns related to editorial control, credibility, and ethical use. This study examined how AI technologies are being integrated into news production and distribution in two mainstream media houses in Kenya: Royal Media Services (RMS) and the Kenya Broadcasting Corporation (KBC). Guided by the Diffusion of Innovation Theory, Unified Theory of Acceptance and Use of Technology, and Technological Determinism Theory, the study explored three research questions: How has AI been integrated into news production processes in RMS and KBC? How has AI been integrated into news distribution processes in RMS and KBC? What challenges hinder the integration of AI technologies in RMS and KBC? Methodologically, the study employed the qualitative research approach and the case study research design, utilizing semi-structured interviews with 5 journalists and 1 data specialist from each of the two selected media houses, drawn from a population of 30 journalists and 4 data specialists in the selected media houses. Purposive and snowball sampling methods were used to identify respondents with experience on AI technologies. Data were thematically analyzed through systematic transcription, coding, theme development, and interpretive synthesis. Ethical principles, including informed consent, confidentiality, and voluntary participation, were observed throughout the study. Findings show that AI has been partly integrated into various processes of news production, particularly content creation, quality enhancement, content curation, and editorial efficiency. In distribution, AI supports audience segmentation, personalized content recommendations, cross-platform optimization, and automated content sharing. However, a full and seamless integration remains constrained by several challenges. These include credibility concerns arising from misinformation and “AI hallucinations,” financial limitations that hinder access to advanced tools, limited AI literacy and training, regulatory uncertainty, data privacy concerns, ethical dilemmas around AI generated content, and resistance from journalists worried about job displacement or loss of editorial autonomy. The study concludes that while AI use in Kenyan newsrooms is growing, it remains uneven and shaped by contextual, technical, and organizational limitations. Media houses should therefore expand AI training, strengthen editorial oversight, invest in cost effective AI solutions, and develop clear editorial guidelines. National regulatory bodies should also provide policy direction to guide responsible and transparent AI adoption and integration in the media sector