Foundations of a CPS Resilience - October 2020

PI: Xenofon Koutsoukos

HARD PROBLEM(S) ADDRESSED

The goals of this project are to develop the principles and methods for designing and analyzing resilient CPS architectures that deliver required service in the face of compromised components. A fundamental challenge is to understand the basic tenets of CPS resilience and how they can be used in developing resilient architectures. The primary hard problem addressed is resilient architectures. In addition, the work addresses scalability and composability as well as metrics and evaluation.

PUBLICATIONS

[1]    Mudassir Shabbir, Jiani Li, Waseem Abbas, Xenofon Koutsoukos. "Resilient Vector Consensus in Multi-Agent Networks Using Centerpoints", American Control Conference 2020. July, 1-3, 2020.
[2]    Waseem Abbas, Mudassir Shabbir, Hassan Jaleel, Xenofon Koutsoukos, "Improving Network Robustness through Edge Augmentation While Preserving Strong Structural Controllability", American Control Conference 2020. July, 1-3, 2020.
[3]    Jiani Li, Waseem Abbas, Mudassir Shabbir, and Xenofon Koutsoukos. “Resilient Distributed Diffusion for Multi-Robot Systems Using Centerpoint”, Robotics: Science and Systems. July 12-16, 2020.
[4]    Himanshu Neema, Xenofon Koutsoukos, Bradley Potteiger, Cheeyee Tang, and Keith Stouffer. “Simulation Testbed for Railway Infrastructure Security and Resilience Evaluation”, Hot Topics in the Science of Security (HotSoS 2020). September 22-24, 2020. Best paper award.
[5]    Ali Ozdagli, Carlos Barreto and Xenofon Koutsoukos. “@PAD: Adverserial Training of Power Systems Against Denial-of-Service Attacks”, Hot Topics in the Science of Security (HotSoS 2020). September 22-24, 2020.
[6]    Jiani Li, Waseem Abbas, Mudassir Shabbir, and Xenofon Koutsoukos. " Resilient Multi-Robot Target Pursuit", Hot Topics in the Science of Security (HotSoS 2020). September 22-24, 2020.
[7]    Scott Eisele, Carlos Barreto, Abhishek Dubey, Xenofon Koutsoukos, Taha Eghtesad, Aron Laszka, Anastasia Mavridou. “Blockchains for Transactive Energy Systems: Opportunities, Challenges, and Approaches”, IEEE Computer, Special Issue on Blockchain & Cyber-physical Systems. vol. 53, no. 9, pp. 66-76, Sept. 2020.

KEY HIGHLIGHTS

This quarterly report presents two key highlights that demonstrate (1) resilient distributed diffusion for multi-robot systems using centerpoint and (2) adversarial training of power systems against denial-of-service attacks.

Highlight 1: Resilient Distributed Diffusion for Multi-Robot Systems Using Centerpoint

In this work, we study the resilient diffusion problem in a network of robots aiming to perform a task by optimizing a global cost function in a cooperative manner. In distributed diffusion, robots combine the information collected from their local neighbors and incorporate this aggregated information to update their states. If some robots are adversarial, this cooperation can disrupt the convergence of robots to the desired state. We propose a resilient aggregation rule based on the notion of centerpoint, which is a generalization of the median in the higher dimensional Euclidean space. Robots exchange their d-dimensional state vectors with neighbors. We show that if a normal robot implements the centerpoint-based aggregation rule and has n neighbors, of which at most ⌈nd+1⌉−1 are adversarial, then the aggregated state always lies in the convex hull of the states of the normal neighbors of the robot. Consequently, all normal robots implementing the distributed diffusion algorithm converge resiliently to the true target state. We also show that commonly used aggregation rules based on the coordinate-wise median and geometric median are, in fact, not resilient to certain attacks. We numerically evaluate our results on mobile multi-robot networks and demonstrate the cases where diffusion with the weighted average, coordinate-wise median, and geometric median-based aggregation rules fail to converge to the true target state, whereas diffusion with the centerpoint-based rule is resilient in the same scenario. Our results are reported in [1].

[1]    Jiani Li, Waseem Abbas, Mudassir Shabbir, and Xenofon Koutsoukos. “Resilient Distributed Diffusion for Multi-Robot Systems Using Centerpoint”, Robotics: Science and Systems. July 12-16, 2020.

Highlight 2: Adversarial Training of Power Systems Against Denial-of-Service Attacks
In this work, we study the vulnerabilities of protection systems that can detect cyber-attacks in power grid systems. We show that machine learning-based discriminators are not resilient against Denial-of-Service (DoS) attacks. In particular, we demonstrate that an adversarial actor can launch DoS attacks on specific sensors, render their measurements useless and cause the attack detector to classify a more sophisticated cyber-attack as a normal event. As a result of this, the system operator may fail to take action against attack-related faults leading to a decrease in the operation performance. To realize a DoS attack, we present an optimization problem to determine which sensors to attack within a given budget such that the existing classifier can be deceived. For linear classifiers, this optimization problem can be formulated as a mixed-integer linear programming problem. In this paper, we extend this optimization problem to find attacks for more complex classifiers such as neural networks. We demonstrate that a neural network, in particular, with RELU activation functions, can be represented as a set of logic formulas using Disjunctive Normal Form, and the optimization problem can be used to efficiently compute a DoS attack. In addition, we propose a defense model that improves the resilience of neural networks against DoS through adversarial training. Finally, we evaluate the efficiency of the approach using a dataset for classification in power systems. Our results are reported in [2].

[2]    Ali Ozdagli, Carlos Barreto and Xenofon Koutsoukos. “@PAD: Adverserial Training of Power Systems Against Denial-of-Service Attacks”, Hot Topics in the Science of Security (HotSoS 2020). September 22-24, 2020.

COMMUNITY ENGAGEMENTS

• Our research was presented in the following conferences: American Control Conference 2020, Robotics: Science and Systems (RSS 2020), Hot Topics in the Science of Security (HotSoS 2020).