Title: Learning from electrical energy market data
Speaker: Prof. Vassilis Kekatos ( Dept. of ECE)
Abstract: The smart grid vision entails advanced information technology and data analytics to enhance the sustainability, security, and economics of the power grid infrastructure. Advances in large-scale algorithms, low-rank and sparse modeling, and online optimization, are definitely pertinent in processing data related to power engineering applications. To this end, this talk presents modern statistical learning tools for inference in electrical energy markets. As a cyber-security example, we will first study the potential of revealing the topology of the underlying physical grid by simply processing real-time energy market prices. As a learning example, we will subsequently exploit the market clearing process and cast day-ahead price forecasting as a low-rank kernel learning problem.
Title: Modeling Stealthy and Persistent Cyber Attacks and Defense Strategies
Speaker: Sudip Saha
Abstract: Persistent and zero-day attacks have increased considerably in the recent past in terms of scale and impact. System owners can no longer rely only on known defenses and thereby protect their resources permanently. It is increasingly common now to observe attackers being able to repeatedly compromise system resources exploiting new vulnerabilities. In this presentation, I will talk about a game theoretic framework to reason about the strategic interaction between the defenders and the attackers in these scenarios. I will discuss two models proposed in the literature for stealthy takeover of systems consisting of single and multiple resources. Then I will talk about a generalized model for networked systems that we are working on and show some preliminary results.