AI / ML Security
Threats and vulnerabilities in learning systems.
Ntung Ngela Landon
I study the security, robustness, and reliability of machine-learning systems, with the goal of building intelligent security systems that stay reliable under adversarial pressure, distribution shift, and evolving threats in malware defense, authentication, and other security-critical settings.
Research Associate · Carnegie Mellon University
View CV
Current research
Malware Defense & Adversarial Robustness
I investigate whether machine-learning-based malware defenses remain effective when attackers modify real-world malware samples. My current Microsoft-funded research involves collecting malware, generating mutations, executing samples in isolated virtualized environments, and evaluating detection failures to identify weaknesses and improve defensive robustness.
Research questions
- How effectively do existing defenses detect mutated malware?
- Which modifications cause detection failures?
- How can defenses be hardened against adaptive evasion?
- How should ML-based security systems be evaluated against evolving threats?
Biometric Security
Face PAD and authentication reliability.
Fairness & Robustness
Evaluation across populations and shifts.
Malware Analysis
Automated analysis and detection systems.
Research statement
My work examines how machine-learning-based security systems behave under adversarial pressure, distribution shift, and changing deployment conditions. I study how evaluation choices, model architecture, datasets, and random variation shape conclusions about security, fairness, robustness, and real-world reliability.
Selected research
Fairness under distribution shift in face PAD
Cross-dataset evaluation, ablations, and multi-seed replication.
Publications
4 research works
Journal article, conference paper, and manuscripts accepted or under review.
Scholarly links
Research
Security, fairness, robustness, and evaluation of AI-enabled systems.
Fairness-aware biometric security
Lightweight face PAD methods for measuring and reducing demographic disparities.
Combined Local Binary Patterns, ethnicity-aware preprocessing, group-specific threshold optimization, and statistical fairness analysis for African and East Asian populations.
Architecture and demographic bias
Comparative evaluation of convolutional and Vision Transformer architectures.
Evaluated ViT-Tiny, ResNet18, and pretrained DeiT-S across African, East Asian, and zero-shot Central Asian groups using APCER, BPCER, ACER, accuracy, and EER.
Robustness and reproducibility
Fairness evaluation under dataset shift, ablations, and random-seed variation.
Applied cross-dataset SSG and LODO protocols, five-seed replication, controlled ablations, and statistical analysis to separate stable model behavior from experimental variability.
Malware mutation and defense evasion
Testing whether malware defenses still detect mutated real-world samples.
In this project, we collect real-world malware samples, mutate them, and run them in virtualized operating system environments to test whether existing malware defense systems can still detect them. The goal is to identify where these defenses fail and propose ways to harden them so that mutated malware cannot evade detection.
Projects
MSc final practicum · CMU
Cross-Platform Automated Malware Analysis Pipeline
Automated Windows and Linux malware routing, isolated analysis, structured reporting, and SOC/SIEM ingestion for the CMU-Upanzi network.
CAPE SandboxPythonVirtual machinesSIEM
Applied cybersecurity
Ethical Hacking & Penetration Testing
Network vulnerability discovery, Wi-Fi security assessment, traffic analysis, and automated vulnerability scanning.
Kali LinuxNmapBurp SuiteWireshark
BEng final project
IoT-Based Fetal Health Monitoring System
Embedded monitoring system for fetal movement with signal filtering and wireless transmission to remote monitoring stations.
IoTSignal processingArduino
Publications
Published work, preprints, and current manuscripts.
Fairness-Aware Face Presentation Attack Detection Using Local Binary Patterns: Bridging Skin Tone Bias in Biometric Systems
J. D. Ndibwile, N. N. Landon, F. Tuyisenge · Journal of Cybersecurity and Privacy, 6(1)
Architectural Bias in Face Presentation Attack Detection: A Comparative Study of Vision Transformers and Convolutional Neural Networks
N. N. Landon, F. Tuyisenge, J. D. Ndibwile · CSP 2026 · arXiv:2606.18510
Fairness Under Distribution Shift in Face Presentation Attack Detection: A Cross-Dataset and Ablation-Based Analysis
N. N. Landon, F. Tuyisenge, R. Dukundane, E. Iduh, J. D. Ndibwile · IEEE Access
Adversarial Artificial Intelligence Threats in Smart Energy Grids: When Learning Systems Learn from the Attacker
J. D. Ndibwile, N. N. Landon, L. Mwinuka · Energy Reports
Investigating how attackers can exploit learning-based systems in smart energy environments and the implications for resilient AI-enabled infrastructure.
Research presentation
Architectural Bias in Face PAD · International Conference on Cryptography, Security and Privacy 2026 · Sapporo, Japan
Experience
Research, applied cybersecurity, teaching, and technical systems work.
Research Associate
Carnegie Mellon University Africa · Kigali, Rwanda
Leads fairness robustness research in face PAD under distribution shift and contributes to a Microsoft-funded malware analysis project.
Research Assistant
Carnegie Mellon University Africa · Kigali, Rwanda
Led comparative ViT, ResNet, and DeiT experiments on architecture, demographic performance, and cross-demographic generalization.
AI, Data Management & Cybersecurity Intern
International Telecommunication Union · Geneva
Investigated platform vulnerabilities and the relationship between AI, data management, and security in AI-enabled information systems.
Research Intern
Carnegie Mellon University Africa · Kigali, Rwanda
Initiated demographic fairness research for underrepresented African populations and developed a lightweight face PAD pipeline.
Teaching Assistant · CMU Africa Bridge Program
Carnegie Mellon University Africa · Kigali, Rwanda
Supported undergraduate students preparing for graduate-level study in computing and technology.
Networks & Telecommunications Intern
Resotel · Douala
Supported surveillance systems, optical-fiber maintenance, secure NAS deployment, and organizational communications infrastructure.
Education
Academic preparation in cybersecurity and computer engineering.
MSc, Information Technology — Cybersecurity
Carnegie Mellon University · 2024–2025
BEng, Computer Engineering
University of Buea · 2019–2023
Contact
For research conversations, collaboration, or speaking opportunities.