Publi­cations

Current state-of-the-art large language models perform exceptionally well for high-resource languages but often lack accuracy for low-resource African languages such as Igbo and Nigerian Pidgin. This study creates benchmark datasets for Igbo-English and Nigerian Pidgin-English translation tasks and evaluates GPT-based models alongside other state-of-the-art translation models specifically designed for low-resource languages.

A collaborative learning framework where ensembles of student GNNs mutually teach each other throughout the training process, enabling knowledge distillation for deploying high-performance graph neural network models in resource-constrained devices.

This paper explores the application of Large Language Models to network intrusion detection in IoT settings, an area that remains largely unexplored despite LLMs gaining significant traction. We fine-tune GPT-3.5 and develop an embedding similarity technique using the CIC IoT 2023 dataset.

Addresses the disconnect regarding the choice and implementation of simulators for resource-constrained devices, proposing a simulation framework for evaluating load sharing strategies that move computation closer to edge devices to minimize network latency.

Employs ChatGPT as an AI curation assistant for curating biomedical literature for population groups, qualitatively and quantitatively evaluating the performance of GPT-4 in curating population information from biomedical literature.

Named Data Networking (NDN), an emerging Internet architecture, alters the basics of the networking model by making content directly addressable and routable. This paper presents the concepts of NDN architecture and a comprehensive overview of emerging application use cases for secure data communications.

Proposes IoT-MGSec, a novel solution to mitigating man-in-the-middle attacks using graph-based learning, which employs graph modeling and embedding techniques to learn node and edge features and generate a robust classifier to detect MiTM attacks with high accuracy.

Addresses privacy concerns in intrusion detection for the Internet of Vehicles by applying split learning — a technique that distributes model training across IoV devices and a central server without sharing raw data — achieving effective intrusion detection while preserving privacy.

Machine translation is an active area of research that has received significant attention over the past decade. In this paper, we survey the current state of the art of machine translation research on Nigerian languages with a major emphasis on neural machine translation techniques.