Roll forward validation based decision tree classification for detecting data integrity attacks in industrial internet of things

Abstract

Data Integrity attack is a major hindrance to the evolution of Industrial Internet of Things (IIoT) as it leads to immense financial loss or even human fatality. The existing security features in Software Defined Networking (SDN), which is emphatically superior to the traditional networks mitigate the integrity attacks to some extent. However, a generic, robust, secure and resilient Intrusion Detection System (IDS) for IIoT is still lacking in the literature. Towards this goal, a generic IDS is already proposed in our earlier research work which combines both anomaly as well as rule-based intrusion detection techniques and successfully tested against the real-time dataset obtained from the water purification process in a test bed at the Singapore University of Technology and Design (SUTD). This research work proposes a supervised learning approach that utilizes Roll-forward technique for validation and Classification and Regression Trees (CART) with invariants for categorization to find anomalousness in the water treatment process. The proposed work incorporates the capability to substantiate time-series data through Roll-forward validation which is then succeeded by utilization of the CART with invariants for classification. The proposed work is simulated using Mininet tool and the train and test accuracies are 99.9% and 98.1% respectively.