Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  
Networked Control Systems Group

Area surveillance using multiple cameras

We study the problem of surveillance and multi-target tracking using multiple cameras. The cameras are of Pan-Tilt (PT) or Pan-Tilt-Zoom (PTZ) and they are mounted at given locations in the environment. The problem is to design an optimization-based controller that is predictive in nature, in the sense that it takes into account the possible future motion of the targets in the optimization problem. The areas of coverage of the cameras may overlap, and as such it becomes necessary to hand off targets from one camera to another in some optimal or suboptimal fashion. We consider the 3 dimensional (3D) geometry of the problem and take into account the nonlinearities that result from the rotation between reference frames as well as the projection onto the image view of the cameras. The underlying optimization problem takes into account the effects of the geometry, the actuation constraints of the cameras, and the maximal possible speed of the targets being tracked. We are also considering hand-offs among the cameras as a scheduling problem over the optimization horizon. The local controller for each camera implements the tracking of multiple targets via a priority-based optimization problem. The resulting optimization problem is naturally a mixed-integer program, which can be implemented in Matlab. An illustration of the geometric aspects of this problem for a single camera and two targets is show in Fig. 1 below.

cameraview
Fig. 1: A schematic of a single camera view in 3D




Cyber attacks in power networks

In large-scale electric power systems, IT infrastructures that provide system-wide supervision and control set-points have been already developed. These systems, known as Supervisory Control and Data Acquisition (SCADA) systems, collect data from remote facilities and send back control commands to the power system. The dependence of power systems on this infrastructure makes them susceptible not only to operational errors but also to external cyber attacks. Our goals are to

  • investigate the impact of such cyber attacks on a large 40-bus network monitored by a SCADA system,
  • determine the vulnerabilities of the interconnected cyber-physical model, and
  • specify the possible attack paths and construct efficient mitigation strategies so as to restore the system and/or avoid cascading failures



twoplants
Fig. 2: Two-area power system with Automatic Generation Control (AGC)


As a first step, we studied the safety of a lumped two-area power system (Fig. 2), and developed a methodology to identify and evaluate the impact that a cyber attack on the Automatic Generation Control (AGC) might have. The primary objective of the AGC, which is one of the few control loops that are closed over the SCADA system without human operator intervention, is to regulate the frequency of each area to the nominal value and maintain the power exchange between the networked two-areas system at a specified level.

We study the impact that such an attack could have by using reachability methods. This way we can detect if a potential attacker is able to construct a policy so as to irreversibly disturb the system. As such, we are interested in providing answers to the following problems:

  • Is the attacker able to steer at least one of the frequencies of the two areas outside of a given safe region?
  • Could the power exchange limits between the two areas be violated?
  • What is the smallest bound for the input of the attacker so as to achieve his goal, i.e. disturb the system?



attacklevels
Fig. 3: Volume of the safe set for different bounds of the attack signal


Fig. 3 shows a family of curves that correspond to different bounds of the attack signal. Based on this data, 200 MW is the minimum signal that an attacker could use in order to disturb the system. Although the frequency bounds in this simple system can not be violated, the attacker is able to construct a policy so as to violate the power exchange limits between the two areas infinitely often, and hence lead to power swinging (Fig. 4).

twoplants
Fig. 4: An example of a destructive attack policy


This would activate the so called out-of-step protection relays, which in turn would trip the generators. Our next step would be to investigate the safety of larger power networks and the effects of cyber attacks on them.

Security of Cyber-Physical Systems & Critical Infrastructures

Cyber-physical systems are ubiquitous in industrial automation systems and critical infrastructures, such as power generation, water and gas distribution, as well as transportation networks. The integration of cyber technologies with physical processes increases system efficiency but, at the same time, it introduces vulnerabilities that undermine the reliability of critical infrastructures. Some infamous examples include the StuxNet computer worm, the Maroochy water breach, and the SQL Slammer worm attack on the Davis-Besse nuclear plant. Thus, cyber-physical systems need to operate reliably in face of unforeseen failures and external malicious attacks. In our work, we propose a mathematical framework for cyber-physical systems, malicious attackers, and security monitors. We characterize fundamental monitoring limitations from system-theoretic and graph-theoretic perspectives. We design centralized and distributed attack detection and identification schemes. We provide insight into the design of coordinated attacks amenable to simple implementation, and we validate our findings through smart grid applications [1] and [2].