Abstract:
This thesis aims to address the critical issue of detecting intrusions in IoT-enabled
critical infrastructure, with a specific focus on water treatment and distribution
plants. The research employs Random forest, SVM and Neural Networks, to
identify the optimal solution for intrusion detection. The models were trained
and tested using a large-scale time series dataset comprising approximately 400k
records. Intrusions were identified based on the time series data, and the stage
of each attack on the test bed was determined. The results indicate that bucket
approach used for training the models was highly effective compared to direct
feed and successfully detects the vast majority of the attacks with a low false
positive rate thus improving on previous works based on this data set. Amongst
all the models compared, neural network outperforms other models in terms of
accuracy and evaluation metrics while using the bucket approach. This study has
the potential to contribute significantly to the field of cyber security for critical
infrastructure systems, particularly for water treatment and distribution plants.
