Industrial Control System (ICS) in this case Supervisory and Data Acquisition (SCADA) is an important role in industry by providing automation processes, centralized control and monitoring processes. SCADA is designed for closed areas with special protocols isolated from the internet using firewalls, but modern SCADA systems are required to be connected to one or more other network protocols to make it easier to access the SCADA system from heterogeneous networks, thereby increasing vulnerability in the SCADA system which causes increased vulnerability to SCADA network system. This phenomenon is interesting to study because there are still few researchers conducting a comprehensive discussion of security on SCADA networks, besides that the performance of open source IDS such as Snort and Suricata is less efficient in detecting disturbances on SCADA networks so that the author has a goal to design an ideal IDS for SCADA networks by using a machine learning approach. Based on the results of the IDS study with machine learning it has good performance from the author’s tests of 5 (five) classification algorithms for machine learning, namely SVM, NAM, DT, RF and KNN with 8 (eight) selected features for detecting activity of port scan, testfr, startdt and invalid causetx produce better KNN performance when compared to the other 4 (four) algorithms that use with the results of the train data accuracy of 98.81% at k = 6. However, the dataset used in this research did not have relevan data for normal and attack scenarios so that in the next study the author will make a testbed to produce more diverse data which results in increased accuracy and reliability of the IDS when implementation.

A Novel Dataset for Experimentation with Intrusion Detection Systems in SCADA Networks using IEC 60870-5-104 Standard

Kreator & Penulis