Sherlock Dataset for Process-aware Intrusion Detection

This document is a dataset paper that introduces Sherlock, a dataset specifically designed for process-aware and network-based intrusion detection in power grid networks. The paper outlines the challenges posed by cyberattacks on critical infrastructures, particularly power grids, which have been increasingly targeted by malicious actors. It details the creation of the Sherlock dataset using the Wattson co-simulator, which emulates realistic network traffic and power grid operations. The dataset encompasses three distinct scenarios featuring various attack types that manipulate process states through malicious commands. The authors evaluate five intrusion detection systems on this dataset, identifying specific challenges that these systems face in the context of power grid networks. The paper emphasizes the need for high-quality datasets in this research area, highlighting the gaps in existing datasets and the importance of Sherlock in advancing the study of intrusion detection methods. The dataset and its documentation are made available for further research and evaluation.