Main Article Content
Abstract
Data breaches against banks, tertiary institutions, hospitals, telecommunication operators, and government agencies in Enugu have grown in frequency and cost, while most organisations in the metropolis still depend on signature-based intrusion detection that cannot recognise novel attacks. This study developed and tested a hybrid artificial intelligence (AI) intrusion detection prototype that combines gradient-boosted classification, autoencoder-based anomaly detection, and a long short-term memory (LSTM) predictive analytics module, and it examined how AI-driven security capabilities relate to breach prevention effectiveness among information technology (IT) personnel in Enugu. A two-phase quantitative design was adopted. In Phase 1, a corpus of 200,000 labelled network flows (120,000 from the CIC-IDS2017 benchmark and 80,000 captured from three pilot organisations in Enugu) was used to train and test six detectors on a stratified hold-out set of 40,000 flows. In Phase 2, a proportionate stratified random sample of 206 IT and cybersecurity personnel drawn from 38 organisations completed a validated 22-item questionnaire. The hybrid prototype achieved 98.54% accuracy, an F1 score of 97.07%, and a false positive rate of 0.89%, outperforming the signature-based baseline (accuracy = 90.66%, F1 = 79.88%); a McNemar test confirmed the difference, χ²(1, N = 40,000) = 2,774.54, p < .001, Cohen’s g = .44. Multiple regression showed that AI intrusion detection capability (β = .34, p < .001, f² = .14), anomaly detection capability (β = .28, p < .001, f² = .09), and predictive analytics capability (β = .21, p = .001, f² = .06) jointly explained 48.5% of the variance in breach prevention effectiveness, F(3, 202) = 63.29, p < .001. The study concludes that layered AI detection substantially strengthens breach prevention in Enugu and recommends phased adoption supported by local traffic data, analyst training, and compliance with the Nigeria Data Protection Act 2023.