Assistant Professor · Artificial Intelligence & Machine Learning · Cybersecurity
University of Cincinnati · ORCID · Google Scholar · DBLP · LinkedIn
I develop and evaluate explainable, efficient, and reliable machine-learning systems. My recent research focuses on cybersecurity, examining how changing conditions, limited resources, and incomplete evidence affect predictions and decisions. I also study explainable decision support in healthcare and education.
My current work is developing around three connected research questions:
How reliable are AI predictions and explanations when operating conditions change?
I study explainability, distribution shift, open-set evaluation, adversarial conditions, and cross-dataset generalization, particularly in malware, ransomware, and intrusion-detection systems.
How do resource constraints and model simplification affect predictive and explanatory reliability?
My work investigates feature refinement, sparsity, lightweight neural architectures, edge inference, and performance–efficiency–robustness trade-offs.
When should an AI system predict, abstain, defer, or require stronger evidence?
This emerging direction investigates evidence sufficiency and integrity, selective prediction, uncertainty-aware decision-making, and human oversight in consequential AI systems.
| Year | Publication | Focus |
|---|---|---|
| 2026 | SHAP-guided memory-forensic malware detection | Feature refinement, robustness, efficiency, and family-disjoint evaluation |
| 2026 | Sparse neural ransomware detection under distribution shift | Temporal, family-disjoint, and open-set evaluation |
| 2026 | Explainable hybrid phishing intrusion detection | Behavioral and structural indicators, ensemble learning, and XAI |
| 2026 | Explainable student-dropout prediction | Dropout prediction across course settings and a proposed intervention framework |
| 2026 | Interpretable breast-tumor classification | Clinical classification with SHAP- and LIME-based interpretation |
Complete publication record: Google Scholar · ORCID · DBLP
My recent work explores evidence-gated decision systems: AI systems that assess whether available evidence is adequate for prediction rather than treating every input as equally decision-ready.
| Work | Status | Focus |
|---|---|---|
| EGAMA-RC | Preprint / emerging research | Selective malware prediction using uncertainty, disagreement, and novelty signals |
| EGAMA-MF | Preprint / emerging research | Checks extraction status, completeness, and provenance before predictive routing |
| Project | Status | Purpose |
|---|---|---|
| S3LiteNet | Benchmarking completed; custom architecture in development | Lightweight and sparse neural models for resource-constrained edge anomaly detection |
| 3-Tier-LSTM | Research code associated with a published study | Experimental pipeline for interpretable breast-tumor classification |
| SynthDataHub | Research and teaching resource | Synthetic datasets for machine-learning experimentation and education |
| Secure Data Analytics with GenAI | Open educational/research resource | Privacy-conscious workflows for AI-assisted analytics and human validation |
I investigate interpretable machine learning and responsible AI for clinical classification and decision support, with attention to model transparency and the relationship between predictive performance and explanation.
I study explainable learning analytics, student-risk prediction, and responsible intervention design, including how predictive evidence can inform rather than replace human judgment.
My teaching and mentorship emphasize experimental rigor, reproducibility, security, responsible AI, and project-based learning.
| Resource | Focus |
|---|---|
| Information Security & Assurance Labs | Applied cryptography, TLS, SSH, VPNs, and information assurance |
| Introduction to Python for IT | Programming foundations and computational thinking |
| Data Technologies Administration Labs | Data architecture, administration, governance, protection, and defense |
| Systematic & Bibliometric Literature Review | Evidence synthesis, bibliometric analysis, and research methods |
I welcome collaboration with researchers, students, and practitioners working on explainable AI, robust machine learning, cybersecurity analytics, distribution shift, edge intelligence, evidence-aware decision systems, and responsible AI deployment.
A question increasingly connecting my work is:
When should an AI system's prediction, and its explanation be trusted enough to support a consequential decision?
For research collaboration or student mentorship, visit my University of Cincinnati research profile or connect with me on LinkedIn.
Academic identifiers
- Full name: Isaac Kofi Nti
- Position: Assistant Professor, School of Information Technology, University of Cincinnati
- ORCID: 0000-0001-9257-4295
- Scopus Author ID: 57210637914
- Web of Science ResearcherID: E-2004-2017
- DBLP: 260/5204
- Institutional profile: University of Cincinnati
Explainable AI · Reliable Decision Systems · Efficient Machine Learning · Cybersecurity