NeuroPAD: A Neural Process-level Anomaly Detection for Industrial Control Systems

点击翻译按钮获取中文摘要

Industrial control system (ICS) refers to the information system used to monitor and control the industrial processes such as manufacturing, production and distribution. While early ICSs were physically isolated from the outside, today's ICSs have evolved to become integrated with the public Internet. Unfortunately, connecting ICSs to Internet creates a multitude of novel threats and attacks against them, as demonstrated in several recent high-profile incidents around the world. Conventional off-the-shelf IT security solutions are ill-suited for defending against stealthy attacks targeting ICS. To defend against this new threat landscape, novel approaches for attack prevention and detection that are tailored to ICS are necessary.The NeuroPAD project is about process-level anomaly detection in ICS. The hypothesis of the NeuroPAD project is that the physical process controlled by the ICS can be modelled by its inputs and outputs to confirm that the control commands sent to the field devices are executed correctly and that the information coming from sensors is consistent with the expected behavior of the system. It proposes to achieve this goal by developing purely data-driven approaches, in which the dynamics of the physical process are learnt from direct system observations using artificial neural networks. Industrial control systems are vital to the operation of Canada's critical infrastructure. The NeuroPAD project is expected to generate new theoretical insights as well as engineering guidelines to serve as foundation for innovative designs in ICS cybersecurity products. Not only this will lead to economic impact by targeting an expanding market but also societal impact by creating solutions that could be used to protect vital critical infrastructures and industrial processes that impact lives and wellbeing of Canadians across the country.