Journal of ICT Systems

Assessment of Vulnerabilities in Student Records Web-Based Systems for Public and Private Higher Learning Institutions in Tanzania

Wilbard G. Masue — 2024-08-23

In spite that HLIs in Tanzania use web-based systems for managing, storing and processing of HLIs information and data such as website contents, academic results and financial records. The HLIs web-based system have been compromised by attackers due to presence of vulnerabilities. The main objective of this study is to assess the vulnerabilities of Students Records Web-based Systems (SRWBS) for private and public Higher Learning Institutions (HLIs) in Tanzania using black-box testing methodology by employing two automatic vulnerability scanners namely OWASP ZAP (Open Webs Application Security Project Zed Attack Proxy; open-source tool) and Acunetix (proprietary tool). This study assesses the vulnerability of SRWBS for 3 private HLIs and 5 public HLIs in Tanzania. The results reveal the total of 29 vulnerabilities which include but are not limited to Broken Authentication and Session Management, Broken Access Control, Security Misconfiguration, Sensitive Data Exposure, Vulnerable JS (Java Script) Libraries, CSRF (Cros Site Request Forgery), Using Components with Known Vulnerabilities, XSS (Cross Site Script), DOM (Document Object Model) based XSS and Reflected XSS. SRWBS of public HLIs were found more vulnerable by average 44.2% than the SRWBS of private HLIs which were vulnerable by average of 37%. Based on these results, this study provides some recommendations for mitigating vulnerabilities and improving the security of SRWBS for private and public HLIs in Tanzania.

Advancing Face Recognition Technologies: The Role of Decision Trees in Classifying Complex Image Pairs

Francis Kakula — 2024-09-25

The advancement of face recognition technologies has been pivotal in various applications, from security systems to personalized user experiences. There are significant efforts already devoted to solving challenges of multimodality and pose variation in face recognition. Some studies focus on multimodality but pose-invariant, and other studies focus on pose variation but single modality. Despite significant progress, various face recognition algorithms do not consider both multimodality and pose variation constraints in their proposed methods. Recognizing face images presented both in a different modality and in a different pose presents serious challenges to current algorithms. This paper proposes an algorithm that combines the strengths of deep learning with decision trees to improve face recognition performance across modalities and poses in constrained and unconstrained environments. This hybrid approach leverages the representational power of deep learning and the interpretability and simplicity of decision trees. The findings indicate significant improvements over existing methodologies, particularly in challenging conditions like when multimodality and pose variation constraints are compounded together in the input face images in both constrained and unconstrained environments. The proposed algorithm not only addresses the limitations of current face recognition systems but also offers scalable, efficient solutions suitable for real-world applications.

Spark: A Statistical Comparison and Evaluation of Classification Algorithms for Fault Prediction in Electrical Secondary Distribution Network

David Makota — 2024-10-30

Managing faults in the electrical secondary distribution network is a challenging task given the nature, size, and complexity. Predicting faults early before they occur helps in increasing the safety and reliability of the power distribution system. Various statistical and machine learning techniques are being used to predict different types of faults. This study applies classification algorithms available in the big data framework Apache Spark through its python interface PySpark to predict electrical secondary distribution network faults. The study evaluates and compares nine algorithms: Decision tree, Gradient-boosted tree, Logistic regression, Naïve Bayes, Multilayer perceptron, Random forest, Linear Support Vector Machine, One-versus-rest and Factorization machines. The research uses Friedman’s test followed by the Nemenyi post hoc test to find the significance of performance differences among the algorithms. The results show significant differences among the algorithms. Gradient-boosted tree and One-versus-rest with Gradient-boosted tree had the best performance for binary and multiclass classification, respectively, while Naïve Bayes had the worst performance.

Forecasting Pediatric Medical Expenses using Machine Learning: A Case Study of the Toto Afya Card in Tanzania

Fransisca Mwakinyali — 2024-12-21

In Tanzania, the discontinuation of the "Toto Afya" insurance package in March 2023 highlighted the challenges of managing escalating medical costs and ensuring the sustainability of child health insurance schemes. This research aimed to predict children's medical expenses using machine learning (ML) algorithms to aid the National Health Insurance Fund (NHIF) in effective cost management. Data from 1,212 registered children, comprising 7 variables, was used to train and test four ML models: linear regression, random forest, XGBoost, and CatBoost. CatBoost emerged as the best-performing model with an accuracy score of 82.1%, followed by XGBoost (79.2%), Random Forest (77.3%), and Linear Regression (76.4%). Additionally, CatBoost and XGBoost achieved the highest F1 scores at 81%. These results underscore CatBoost's effectiveness in accurately predicting children's medical expenses, which is crucial for NHIF's cost management and sustainability of insurance schemes. By leveraging ML algorithms like CatBoost, NHIF can improve evidence-based decision-making for policymakers and healthcare providers, ultimately enhancing access to affordable healthcare for children. This approach supports NHIF's goals and contributes to better healthcare outcomes and sustainability.