The Invisible Shield: Privacy-Preserving Federated AI for Detecting Cyberattacks in IoT Networks

Primary supervisor

Sanoop Mallissery

Please note that this Honours and Masters Project topic is offered exclusively at our Monash Malaysia campus and is not available at the Clayton campus.

This project is not just another IDS project. It sits at the intersection of four powerful areas:

So you will deal with:

Cybersecurity: detecting attacks in IoT and IIoT networks.
Federated Learning: training AI without centralizing raw data.
Privacy Engineering: measuring and reducing leakage from model updates.
Edge AI: making the system lightweight enough for constrained devices.

Most Master’s projects stop at “my model achieved high accuracy.” This project goes beyond that by asking:

Can the model protect privacy?
Can it survive malicious clients?
Can it work under non-IID real-world IoT conditions?
Can it run efficiently at the edge?
Can we prove the privacy-security-efficiency trade-off experimentally?

Aim/outline

We aim to build a privacy-preserving Federated Learning-based Intrusion Detection System for IoT/IIoT networks. Instead of sending raw network traffic to a central server, multiple IoT clients collaboratively train an AI model while keeping their data local.

The project will explore one key question:

Can IoT devices learn together to detect cyberattacks without exposing private data?

So you will be developing a prototype that includes:

  1. IoT/IIoT dataset preparation using public IDS datasets.
  2. Non-IID client simulation to represent realistic IoT environments.
  3. Baseline IDS model development using ML/deep learning.
  4. Federated Learning pipeline using methods such as FedAvg/FedProx.
  5. Privacy protection using Differential Privacy or Secure Aggregation-inspired mechanisms.
  6. Optional robustness testing against poisoning/backdoor clients.

• 7. Optional edge-efficiency analysis using pruning, quantization, or communication reduction.

URLs/references

Datasets
Edge-IIoTset: https://ieee-dataport.org/documents/edge-iiotset-new-comprehensive-realistic-cyber-security-dataset-iot-and-iiot-applications
ToN-IoT: https://research.unsw.edu.au/projects/toniot-datasets
Bot-IoT: https://research.unsw.edu.au/projects/bot-iot-dataset
WUSTL-IIOT-2021: https://www.cse.wustl.edu/~jain/iiot2/index.html
CIC IIoT Dataset 2025: https://www.unb.ca/cic/datasets/iiot-dataset-2025.html

Federated Learning
FedAvg: https://proceedings.mlr.press/v54/mcmahan17a.html
FedProx: https://arxiv.org/abs/1812.06127
Federated Learning Survey: https://arxiv.org/abs/1912.04977

Privacy / Security
Differential Privacy Book: https://www.cis.upenn.edu/~aaroth/Papers/privacybook.pdf
Secure Aggregation: https://dl.acm.org/doi/10.1145/3133956.3133982
Deep Leakage from Gradients: https://arxiv.org/abs/1906.08935

Required knowledge

Essential:
Python, machine learning basics, cybersecurity fundamentals, network traffic/IDS concepts, data preprocessing, evaluation metrics such as accuracy, precision, recall, F1-score, and AUC.

Useful:
PyTorch or TensorFlow, Scikit-learn, Pandas, NumPy, Federated Learning frameworks such as Flower/FedML, basic Differential Privacy, Git/GitHub, Linux command line.

Nice to have:
Adversarial machine learning, poisoning/backdoor attacks, Secure Aggregation, model pruning/quantization, non-IID data partitioning, Docker, experiment tracking.