The problem
IoT devices are a favourite target of botnets. A gateway sees only its own traffic and few attacks; sharing traffic is a privacy problem. Detectors must also be small enough to run on a router.
Who needs it
Router and smart-home vendors, ISPs, enterprise network security.
Idea of the approach
- Datasets: UNSW-NB15 and an IoT botnet dataset, split across simulated gateways.
- Lightweight detectors (autoencoder like Kitsune, small MLP), trained locally then federated.
- Measure detection rate, false alarms, model size and inference time on a Raspberry Pi.
What you will build and deliver
- Federated IDS prototype and evaluation.
- Paper.
Skills needed
Python, networking basics, machine learning.
Research contribution & where to publish
Contribution: size-constrained federated IDS with realistic non-IID traffic. Venues: IEEE CNS, IEEE ICC, Computer Networks, IEEE IoT Journal.
Key references — start reading here
- N. Moustafa, J. Slay. UNSW-NB15: a comprehensive data set for network intrusion detection systems (UNSW-NB15 network data set). MilCIS, 2015.
- Y. Mirsky, T. Doitshman, Y. Elovici, A. Shabtai. Kitsune: An Ensemble of Autoencoders for Online Network Intrusion Detection. NDSS, 2018.
- T. D. Nguyen, S. Marchal, M. Miettinen, H. Fereidooni, N. Asokan, A.-R. Sadeghi. DÏoT: A Federated Self-learning Anomaly Detection System for IoT. IEEE ICDCS, 2019.
- B. McMahan, E. Moore, D. Ramage, S. Hampson, B. Agüera y Arcas. Communication-Efficient Learning of Deep Networks from Decentralized Data. AISTATS, 2017.
Interested?
Send the topic title and the courses you have taken.
Contact Dr. Hamza Djigal