Automated bullying-content detection has advanced rapidly for English and other high-resource languages, yet comparable evidence for Swahili remains limited, particularly for short message service (SMS) communication in Tanzania. This study developed and evaluated a context-aware deep learning model for binary classification of bullying and non-bullying Swahili SMS messages. A multi-source corpus of 7,228 messages was initially assembled from prior Swahili datasets, voluntary student contributions, and Google Forms; duplicate records were removed during data cleaning before the train-validation partition was created. Four Kiswahili graduates applied a common annotation framework that considered the target, communicative intent, and surrounding linguistic context rather than treating offensive vocabulary as a sufficient label criterion. The messages were normalised, tokenised with a 10,000-token vocabulary, padded to 100 tokens, and represented using trainable 300-dimensional FastText embeddings. A Bidirectional Long Short-Term Memory network used 160 units in each direction, followed by dropout, a 96-unit rectified linear dense layer with L2 regularisation, and a two-class Softmax output. Candidate configurations were assessed through Keras Tuner and validation-based model selection. On the 1,470-message internal validation set created after duplicate removal, the selected model achieved 90.07% accuracy, 90.09% macro precision, 90.05% macro recall, and 90.06% macro F1-score. The confusion matrix contained 641 true negatives, 683 true positives, 80 false positives, and 66 false negatives. Bullying recall reached 91.19%, indicating that the model identified most harmful messages, although the 8.01-percentage-point training-validation gap and divergent loss curves showed moderate overfitting. The study contributes a Tanzania-focused Swahili SMS resource, an empirically evaluated FastText-BiLSTM architecture, and deployment guidance that positions automated detection as a triage mechanism for human review rather than an autonomous enforcement tool.
Andrea Peter, Gustaph Sanga, G. Tesha· East African Journal of Info...· 0 citations
Mobile money is critical financial infrastructure in Tanzania, but impersonation-based fraud can bypass technical safeguards by inducing users to authorise payments themselves. This study examined the techniques, patterns, and vulnerabilities facilitate impersonation-based fraud and the factors predicting user compliance with fraudulent instructions in Tanzania. A mixed-methods design combined a survey of 231 active mobile money users in Dar es Salaam and Kibaha, eight key informant interviews and a review of regulatory and industry documents. Impersonation attempts were pervasive, with 72.3% of users targeted and 86.2% of those targeted were repeatedly. Attackers requested PIN in only 0.6% of cases but inducing 98.8% to send money themselves. Among targeted users, 74.3% complied, and 97.6% of compliers lost money. Logistic regression identified trust in providers-appearing communications as the only significant unique predictor of compliance (OR = 2.50). Users strongly preferred transaction-context safeguards, with 93.1% wanting cancellation capability and 87.4% wanting more verification time. Impersonation-based mobile money fraud is primarily an authorised-payment problem rooted in trust exploitation. Mitigation should therefore strengthen safeguards at the point of payment authorisation
Gregory Aloyce, C. Budoya, G. Tesha· East African Journal of Info...· 0 citations
Emergency access to Electronic Health Record (EHR) systems presents a difficult balance between protecting sensitive patient information and ensuring that clinicians can obtain critical information without delay. Existing security approaches, including Zero-Trust Architecture (ZTA), multi-factor authentication (MFA), and conventional break-glass mechanisms, primarily rely on user identity, device, or network context and do not consider the patient's current clinical condition. No existing approach uses the patient's real-time clinical deterioration as the signal that drives the access decision, which is the specific gap this study addresses. The novelty of the proposed framework lies in coupling clinical-deterioration prediction directly to Zero-Trust policy enforcement, together with automatic privilege revocation once the patient stabilises, rather than in any single component. This study proposes an AI-driven risk-adaptive Zero-Trust framework that incorporates real-time patient deterioration into access control decisions. An LSTM model analyses five vital signs and classifies patient status as STABLE, WARNING, or CRITICAL. The predicted clinical risk is combined with role-specific emergency authority to calculate a composite risk score that determines one of four access levels: Direct Access, MFA Required, Restricted Access, or Denied. The framework was evaluated using the MIMIC-III Clinical Database Demo, comprising 98 patients and 5,992 NEWS2-labelled time-series sequences, together with 6,000 simulated access requests across six clinical specialities. The proposed model achieved an overall accuracy of 82.20%, a CRITICAL-class recall of 88.00%, and an AUC of 0.9752. The access-control engine produced an average decision latency of 0.265 ms while maintaining complete audit logging and 99.3% least-privilege compliance. These findings suggest that integrating clinical deterioration predictions into Zero-Trust access control can improve emergency responsiveness while preserving security and accountability. Although the framework was evaluated in a simulated environment, the results demonstrate its potential for future deployment and validation in real clinical settings.
Arthur Nashon Malingo, Christian Budoya, G. Tesha· East African Journal of Info...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
