Cybersecurity researcher working on federated learning, intrusion detection, and explainable AI.
Selected publications, newest first.
Case files from research and engineering work.
A robust aggregation scheme for training intrusion detection models across organizational boundaries without centralizing raw traffic data, evaluated against poisoning and non-IID client distributions.
A stacked ensemble for intrusion detection that holds up when the network changes. GRU, LSTM, DNN and MLP base models are each trained on a different NetFlow dataset and then frozen, and a meta-learner combines them, so every domain's decision boundary survives into the final model.
Working miniatures of my research: small, real computations you can poke at, all running in your browser.