Paul J Bonczek

I am currently a Ph.D. candidate in Electrical Engineering at the University of Virginia. I work in the Autonomous Mobile Robots (AMR) Lab and Link Lab under the advisement of Prof. Nicola Bezzo. Before joining AMR Lab, I obtained my B.S. degree dual majoring in Electrical & Computer Engineering and Applied Mathematics at the State University of New York (SUNY) Polytechnic Institute in 2016. 

My research focuses on cyber-physical system cyber-security, resilient control and planning of single- and multi-vehicle autonomous systems, resilient state estimation, run-time attack detection algorithms, and adaptive control systems. 

Bonczek_Paul.jpeg

Research

Monitoring for Non-Randomness Behavior in Residual-based Detection Schemes for Discovery of Stealthy Sensor Attacks

Today's autonomous cyber-physical systems are fitted with multiple on-board sensors and computers, allowing for increased capabilities in various applications to perform complex tasks. Consequently, these enhancements increase the risk of cyber attackers to spoofing attacks that can compromise system integrity and safety. With a growing number of vulnerable entry points for attackers on progressively impactful systems within our society, it is imperative to develop more strict security measures to ensure proper performance.

 

Intelligent attackers are able to implement stealthy attack sequences that hide within system uncertainties in order to remain hidden from typical state-of-the-art detection techniques. However, to effectively hijack a system, an attack must create behavior anomalies (i.e., non-randomness) that contradict the known system model(s). We develop attack detectors that are able to discover non-random events due to sensor cyber-attacks in residual-based detection schemes. With our proposed attack detectors implemented on-board a system, we show that previously undetectable attacks can now be found. More details about our approaches with simulation and experiment results can be found in our papers. [ACC '20], [IFAC '20], [ACC '21].

ACC20_intro.png
System_Architecture_CPSdetection.png

Detection of Non-Random Behavior for Resilient Coordination of Robotic Swarms

TRO_architecture.png

Within robotics, multi-agent system coordination and swarming have long been studied, but have regained consideration attention to consider safety and security issues. Most multi-agent applications are designed without considering cyber-security issues, which perform operations with the assumption that all agents are cooperating. We propose a monitoring framework for multi-agent systems that detects and isolates compromised vehicles to enable the swarm to resiliently continue operations in the presence of stealthy communications and sensor attacks.

 

Additional simulation and experiment results of our resilient multi-agent framework can be found in the following webpage. This work has recently been accepted to the Transactions on Robotics (T-RO) and will be included in an upcoming Special Issue on Resilience in Networked Robotic Systems. To learn more about our framework, the details may be found in our paper [TRO '22].

Detection and Inference of Randomness-based Behavior for Resilient Multi-vehicle Coordinated Operations

Generally, applications that leverage multi-vehicle networks assume that all vehicles are cooperative while performing desired operations to maintain swarming formations and can exchange all necessary information to complete a task. However, these vehicles are susceptible to malicious external cyber-attacks that can affect system performance or intercept safety-critical information. We propose to utilize a side-channel information exchange that contains hidden data for vehicles in the network to detect that is unknown to potential attackers.

 

Virtual spring-damper meshes for proximity-based formation control are leveraged for both detection of misbehaving vehicles and also any hidden signatures that are being passed from nearby vehicles. In our work, we apply our framework to applications where mobile robotic teams relay safety-critical information concerning any discovered object of interest to the remaining vehicles by utilizing hidden signatures to maintain secrecy from potential attackers.

 

More details about our framework can be found in our paper [IROS '21]. Further simulation and experiment videos can be viewed from the following webpage.

IROS21_PAUL_intro.png

Cooperative Recovery for Safe Multi-robot Operations

Our current work focuses on maintaining resilience in multi-robot systems as cyber attacks or faults occur to individual robots within the formation. A single robot that is compromised can degrade the swarm's formation control performance, and potentially cause devastating effects.

 

In our previous work (i.e., T-RO '22), compromised agents within the swarm are isolated and removed from the control network to ensure the multi-robot system can resiliently maintain operations. However, isolated vehicles that lose localization capabilities are left behind by the remaining agents in the swarm. In this work, we propose a framework for neighboring robots to cooperatively aid in re-localization of compromised agents. We propose to leverage a hidden signal-based alerting mechanism to notify neighboring robots of the impending attack/fault. The uncompromised neighbors then behave in a cooperative manner to aid in re-localization to save the compromised agent, then continue to complete the desired operation.

We are currently preparing this work for submission to the IEEE Robotics and Automation Letters (RA-L) in the Summer of '22. Videos will be provided in the near future!

CooperativeRecovery.png

Recovery of Autonomous Systems Operating During On-board Controller Attacks

Cyber attacks, failures, and implementation errors inside of an on-board controller of an autonomous system can affect its correct behavior, which can lead to unsafe states and degraded performance. For example, a scenario may occur when cyber attacks manipulate controller feedback gain parameters or block control inputs to trigger an undesired behavior of the system based on state and reference signal information provided to the controller. In this work, we seek to detect the compromised regions within the information provided to the controller and then design a recovery method to counteract the malicious effects to the controller to allow for resilient operations such that a robot can maintain desired control performance.

This work has recently been accepted to IROS '22 which is taking place in Kyoto, Japan on October 13-17, 2022. To see our simulation and experiment results, visit our webpage for this paper.

IROS22_Intro4.png

Resilient Multi-agent Formation Control via RSSI-based Localization

Cooperative multi-agent systems exchange information to coordinate their motion, such as in maintaining specific formations while performing operations. When localization is compromised on one or more agents, due to scenarios like cyber attacks or faults to on-board positioning sensors, desired formation control cannot be preserved.

To deal with this, we propose to utilize received signal strength indication (RSSI) as an inexpensive method to aid in localization (computationally and weight-wise). In this work, we formalize: i) detection of compromised position sensor measurements, ii) sensor reconfiguration to switch to RSSI-based position measurements, and iii) a robust method to adaptively estimate noise covariance of the RSSI-based sensing when fitted to a Kalman filter. This work is being prepared for submission to the 2023 IEEE American Control Conference (ACC).

ACC23_intro.png

Cooperative Robot Teams for Defending Against Malicious Intruders

Present-day robot teams are being used to accomplish various civilian and military applications, such as  surveillance and security. My most recent project involves the development of robotic teams to collaborate their motion in order to defend against potential disasters. These disasters could include maliciously behaving intruders attempting to interfere with safety-critical infrastructure or regions within an environment. When behaving cooperatively, the multi-robot team can more effectively detect and engage an intruding agent. In this work, we develop a framework for a  robot team to defend a "protected" region within an environment from intruders (i.e., prevent the intruder from entering the region) then, if possible, shepherd the maliciously behaving intruder to a safe region.

As this work progresses, we will provide more details which include pictures and simulation/experiment videos!

Hobbies

Outside of work, I enjoy playing pickleball, cooking, hiking, playing/watching sports, fantasy football, and learning about WWII history.