In today’s connected world, there is a tendency of connectivity even in the sectors which conventionally have been not so connected in the past, such as power systems substations. Substations have seen considerable digitalization of the grid hence, providing much more available insights than before. This has all been possible due to connectivity, digitalization and automation of the power grids. Interestingly, this also means that anybody can access such critical infrastructures from a remote location and gone are the days of physical barriers. The power of connectivity and control makes it a much more challenging task to protect critical industrial control systems. This capability comes at a price, in this case, increasing the risk of potential cyber threats to substations. With all such potential risks, it is important that they can be traced back and attributed to any potential threats to their roots. It is extremely important for a forensic investigation to get credible evidence of any cyber-attack as required by the Daubert standard. Hence, to be able to identify and capture digital artifacts as a result of different attacks, in this paper, the authors have implemented and improvised a forensic testbed by implementing a sandboxing technique in the context of real time-hardware-in-the-loop setup. Newer experiments have been added by emulating the cyber-attacks on WAMPAC applications, and collecting and analyzing captured artifacts. Further, using sandboxing for the first time in such a setup has proven helpful.
Challenges in Machine Learning based approaches for Real-Time Anomaly Detection in Industrial Control Systems
