About · Research
Published Research
Peer-reviewed work from the Harbor Coast University networks lab on intrusion detection, IoT telemetry, and resilient architectures. Full texts are available on request.
Journal of Network Intelligence · 2024
Skew-Aware Federated Aggregation for IoT Intrusion Detection
M. Carter, D. Osei, L. Whitaker — Harbor Coast University Networks Lab
IoT gateways observe radically different traffic, which breaks standard federated averaging and lets minority attack patterns vanish in the merge. We introduce a clustered aggregation scheme that groups gateways by traffic profile before merging updates, preserving detection quality for rare attack families. Experiments on partitioned public intrusion traces show the approach holds steady under severe data skew where baseline methods degrade. The paper details the threat model, communication budget, and gateway-class hardware profile.
Federated LearningIntrusion DetectionIoT Security
Workshop on Edge AI for Communications · 2023
Lightweight Sequence Models for Gateway-Side Threat Detection
M. Carter, L. Whitaker — Harbor Coast University Networks Lab
Running detection on the gateway itself removes the need to backhaul raw traffic, but only if the model fits the hardware. This work compares compact sequence architectures on flow-metadata features, measuring detection quality against CPU, memory, and power budgets on commodity gateway boards. We publish the full benchmarking harness so other teams can repeat the study on their own devices and traffic mixes.
Edge AIBenchmarkingFlow Analysis
Undergraduate Research Symposium · 2023 · Best Paper Nominee
Visualizing Metro Fiber Health: From OTDR Traces to Operator Intuition
M. Carter — Senior Capstone (NetLens), advised by Dr. R. Fontaine
OTDR traces are rich but illegible to anyone who has not spent years reading them. NetLens, my senior capstone, translates trace events into an interactive map of fiber health — reflective faults, bend losses, and degrading splices rendered where crews can act on them. The paper covers the event-classification pipeline, the field validation with two maintenance crews, and why the project earned Best Senior Project.
Fiber OpticsVisualizationOTDR
