Analytical Assessment of the Research Gap in AI-Based Intrusion Detection Systems for Nigerian Network Environments

Machine learning-based intrusion detection systems offer significant performance advantages over signature-based approaches, but their evaluation in the specific traffic patterns and threat profiles of Nigerian institutional environments represents an important research gap. This study systematically examined this gap through scoping review and original empirical analysis. A scoping review of 12 databases identified 47 publications from 2019 to 2024 on ML-based IDS applied to sub-Saharan African network environments, of which only 3 used Nigerian network traffic datasets. Original empirical analysis compared five ML algorithms (Random Forest, SVM, KNN, Decision Tree, and LSTM) on the NSL-KDD benchmark dataset and a supplementary Nigerian bank network traffic capture dataset collected with institutional authorisation. On NSL-KDD, Random Forest achieved 99.2 percent detection accuracy. On the Nigerian traffic dataset, accuracy dropped to 87.4 percent across all algorithms, with false positive rates 3.2 times higher than benchmark results due to distinctive Nigerian network usage patterns not represented in international datasets. The performance degradation was most severe for encrypted financial transaction traffic. The study identifies the absence of labelled Nigerian network traffic datasets as the primary research gap and recommends the establishment of a national academic cybersecurity traffic dataset repository hosted by NITDA to enable Nigerian-context AI security research.

Keywords: AI intrusion detection, machine learning, Nigerian network traffic, research gap, IDS performance