Abstract
The degree of sophistication of modern cyber-attacks has increased in recent years, and in the future these attacks will more and more target cyber-physical systems (CPS). Unfortunately, today’s security solutions that are used for enterprise information technology (IT) infrastructures are not sufficient to protect CPS, which have largely different properties, involve heterogeneous technologies, and have an architecture that is tailored to specific physical processes. The objective of the synERGY project was to develop new methods, tools and processes for cross-layer anomaly detection (AD) to enable the early discovery of both cyber- and physical-attacks with impact on CPS. To this end, synERGY developed novel machine learning approaches to understand a system’s normal behaviour and detect consequences of security issues as deviations from the norm. The solution proposed by synERGY are flexibly adaptable to specific CPS layers, thus improving the detection capabilities. Moreover, synERGY interfaces with various organizational data sources, such as asset databases, configuration management, and risk data to facilitate the semi-automatic interpretation of detected anomalies. The synERGY approach was evaluated in a utility provider’s environment. This paper reports on the general architecture and the specific pitfalls that needed to be solved, during the design, implementation and deployment of the synERGY system. We foresee this work to be of benefit for researchers and practitioners, who design and implement security systems that correlate massive data from computer logs, the network or organizational context sources, to timely detect cyber attacks.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have