May 4, 2017: Invited Speaker Seminar: Joint Attack Detection and Secure State Estimation of Cyber-Physical Systems

Dr. Nicola Forti
Carnegie Mellon University

Thursday, May 4, 2017 at 11:00 am
Room EV003.309

Abstract

The tighter interaction between cyber and physical realms is unavoidably introducing novel security vulnerabilities, which render next-generation Cyber-Physical Systems (CPSs) such as smart grids, vehicular networks, etc. subject to non-standard malicious threats. This is why the design of secure CPSs has become a top priority. In particular, the focus of this talk is on secure state estimation, i.e. on how to reconstruct the state of CPSs even when under different (and possibly combined) types of attacks. By following a Bayesian random set approach, we show how the problem of jointly detecting data integrity attacks and securely estimating the state of the CPS also in presence of fake measurement injections can be formulated and solved. The random set paradigm is used to model the switching nature of both integrity attacks and fake packet injections, while the stochastic Bayesian framework allows to account for several sources of randomness (e.g. process and measurement noises, attack signal, fake measurements) in a probabilistic way unlike deterministic attack monitors based on residual analysis.

The core advantages of this Bayesian approach are: i) it can encompass in a unique and general framework various classes of attacks, ii) can deal with nonlinear systems, iii) takes into account the presence of disturbances and noise, iv) propagates in real-time useful information for dynamic state estimation, attack reconstruction and security decision-making. Unfortunately, the proposed filter does not have, in general, a closed-form solution, but it can be practically implemented as particle or Gaussian-mixture filter. The effectiveness and the real-world applicability of the developed tools are tested on standard power systems. Further advances investigate worst-case performance degradation analysis, the definition of a probabilistic notion of attack stealthiness and the application of this Bayesian approach to secure state estimation in distributed settings.

Biography

Nicola Forti is a Postdoctoral Research Fellow in the Department of Electrical and Computer Engineering at Carnegie Mellon University, Pittsburgh, PA, and in the Department of Information Engineering at the University of Florence, Italy. He received his Ph.D. (summa cum laude) in Information Engineering from the University of Florence, Italy in 2016. He also completed his B.S. in Mechanical Engineering and M.S. (cum laude) in Electrical and Automation Engineering at the University of Florence, Italy in 2013. In 2015 he was Visiting Research Scholar in the Department of Electrical and Computer Engineering at Carnegie Mellon University for one year. 

Dr. Forti's main research interests include networked control systems, data fusion, secure estimation and control of cyber-physical systems and multi-target tracking.