Recent incidents have shown that Industrial Control Systems (ICS) are becoming increasingly susceptible to sophisticated and targeted attacks initiated by adversaries with high motivation, domain knowledge, and resources. Although traditional security mechanisms can be implemented at the IT-infrastructure level of such cyber-physical systems, the community has acknowledged that it is imperative to also monitor the process-level activity, as attacks on ICS may very well influence the physical process. In this paper, we present pasad, a novel stealthy-attack detection mechanism that monitors time series of sensor measurements in real time for structural changes in the process behavior. We demonstrate the effectiveness of our approach through simulations and experiments on data from real systems. Experimental results show that pasad is capable of detecting not only significant deviations in the process behavior, but also subtle attack-indicating changes, significantly raising the bar for strategic adversaries who may attempt to maintain their malicious manipulation within the noise level.
1 September 2020 in International Journal of Critical Infrastructure Protection
One of the main tasks sought after with machine learning is classification. Support vector machines are one of the widely used machine learning algorithms for data classification. SVMs are by default binary classifiers, extending them to multi-class classifiers is a challenging on-going research problem. In this paper, we propose a new approach to constructing the multi-class classification function, where the structure and properties of the support vectors are exploited without altering the training procedure. Our contribution is based on the insight that one is not restricted to using the hyperplane-based decision function, resulting from the mathematical optimization problem. The proposed classification procedure considers the notion of distance between vectors in feature space. We show how, under the assumption of a normalized kernel, the distance between two vectors in feature space can be expressed solely in terms of their inner product. We apply both the original and proposed methods on synthetic datasets in a simulation setting, and then we argue that the proposed distance-based method represents a more rigorous and intuitive measure than the traditional hyperplane-based method.
Incorporating information and communication technology in the operation of the electricity grid is undoubtedly contributing to a more cost-efficient, controllable, and flexible power grid. Although this technology is promoting flexibility and convenience, its integration with the electricity grid is rendering this critical infrastructure inherently vulnerable to cyberattacks that have potential to cause large-scale and far-reaching damage. In light of the growing need for a resilient smart grid, developing suitable security mechanisms has become a pressing matter. In this work, we investigate the effectiveness of a model-free state-of-the-art attack-detection method recently proposed by the cybersecurity community in detecting common types of cyberattacks on voltage control in distribution grids. Experimental results show that, by monitoring raw controller and smart-meter data in real time, it is possible to detect denial of service, replay, and integrity attacks, thus contributing to a resilient and more secure grid.