Automated vehicles are getting closer each day to large-scale deployment. It is expected that self-driving cars will be able to alleviate traffic congestion by safely operating at distances closer than human drivers are capable of and will overall improve traffic throughput. In these conditions, passenger safety and security is of utmost importance. When multiple autonomous cars follow each other on a highway, they will form what is known as a cyber-physical system. In a general setting, there are tools to assess the level of influence a possible attacker can have on such a system, which then describes the level of safety and security. An attacker might attempt to counter the benefits of automation by causing collisions and/or decreasing highway throughput. These strings (platoons) of automated vehicles will rely on control algorithms to maintain required distances from other cars and objects around them. The vehicle dynamics themselves and the controllers used will form the cyber-physical system and its response to an attacker can be assessed in the context of multiple interacting vehicles. While the vehicle dynamics play a pivotal role in the security of this system, the choice of controller can also be leveraged to enhance the safety of such a system. After knowledge of some attacker capabilities, adversarial-aware controllers can be designed to react to the presence of an attacker, adding an extra level of security. This work will attempt to address these issues in vehicular platooning. Firstly, a general analysis concerning the capabilities of possible attacks in terms of control system theory will be presented. Secondly, mitigation strategies to some of these attacks will be discussed. Finally, the results of an experimental validation of these mitigation strategies and their implications will be shown.

Additive manufacturing (AM) is a growing section of the advanced manufacturing field and is being used to fabricate an increasing number of critical components, from aerospace components to medical implants. At the same time, cyber-physical attacks targeting manufacturing systems have continued to rise. For this reason, there is a need to research new techniques and methods to ensure the integrity of parts fabricated on AM systems. This work seeks to address this need by first performing a detailed analysis of vulnerabilities in the AM process chain and how these attack vectors could be used to execute malicious part sabotage attacks. This work demonstrated the ability of an internal void attack on the .STL file to reduce the yield load of a tensile specimen by 14% while escaping detection by operators. To mitigate these vulnerabilities, a new impedance-based approach for in situ monitoring of AM systems was created. Two techniques for implementing this approach were investigated, direct embedding of sensors in AM parts, and the use of an instrumented fixture as a build plate. The ability to detect changes in material as small as 1.38% of the printed volume (53.8 mm3) on a material jetting system was demonstrated. For metal laser powder bed fusion systems, a new method was created for representing side-channel meltpool emissions. This method reduces the quantity of data while remaining sensitive enough to detect changes to the toolpath and process parameters caused by malicious attacks. To enable the SCMS to validate part quality during fabrication required a way to receive baseline part quality information across an air-gap. To accomplish this a new process noise tolerant method of cyber-physical hashing for continuous data sets was presented. This method was coupled with new techniques for the storage, transmission, and reconstructing of the baseline quality data was implemented using stacks of "ghost" QR codes stored in the toolpath to transmit information through the laser position. A technique for storing and transmitting quality information in the toolpath files of parts using acoustic emissions was investigated. The ATTACH (additive toolpath transmission of acoustic cyber-physical hash) method used speed modulation of infill roads in a material extrusion system to generate acoustic tones containing quality information about the part. These modulations were able to be inserted without affecting the build time or requiring additional material and did not affect the quality of the part that contained them. Finally, a framework for the design and implementation of a SCMS for protecting AM systems against malicious cyber-physical part sabotage attacks was created. The IDEAS (Identify, Define, Establish, Aggregate, Secure) framework provides a detailed reference for engineers to use to secure AM systems by leveraging the previous work in vulnerability assessment, creation of new side-channel monitoring techniques, concisely representing quality data, and securely transmitting information to air-gapped systems through physical emissions. / Doctor of Philosophy / Additive manufacturing (AM), more widely known as 3D printing, is a growing field of manufacturing where parts are fabricated by building layers of material on top of each other. This layer-based approach allows the production of parts with complex shapes that cannot be made using more traditional approaches such as machining. This capability allows for great freedom in designing parts, but also means that defects can be created inside of parts during fabrication. This work investigates ways that an adversary might seek to sabotage AM parts through a cyber-physical attack. To prevent attacks seeking to sabotage AM parts several new approaches for security are presented. The first approach uses tiny vibrations to detect changes to part shape or material by attaching a small sensor either directly to the parts or to the surface that they are built on. Because an attack that sabotages an AM system (3D printer) could also affect the systems used to detect part defects these systems should be digitally separated from each other. By using a series of QR codes fabricated by the AM system along with the parts, information can be sent from the AM system to the monitoring system through its sensors. This prevents a cyber-attack from jumping from the AM system to the monitoring system. By temporarily turning off the laser power and tracking the movements of the guiding mirrors the QR code information can be sent to the monitoring system without having to actually print the QR code. The information stored in the QR code is compared to the emission generated when fabricating the parts and is used to detect if an attack has occurred since that would change the emissions from the part, but not from the QR code. Another approach for sending information from the AM system using physical emissions is by using sounds generated during part fabrication. Using a desktop scale 3D printer, the speed of certain movements was increased or decreased. The change in speed causes the sound emitted from the printer to change, while not affecting the actual quality of the print. By using a series of tones, similar to Morse code, information can be sent from the printer. Research was performed on the best settings to use to transmit the information as well as how to automatically receive and decode the information using a microphone. The final step in this work is a framework that serves as a guide for designing and implementing monitoring systems that can detect sabotage attacks on AM parts. The framework covers how to evaluate a system for potential vulnerabilities and how to use this information to choose sensors and data processing techniques to reduce the risk of cyber-physical attacks.

With recent advancements in computer and network technologies, cyber-physical systems have become more susceptible to cyber-attacks; and production systems are no exception. Unlike traditional Information Technology (IT) systems, cyber-physical systems are not limited to attacks aimed at Intellectual Property (IP) theft, but also include attacks that maliciously affect the physical world. In manufacturing, such cyber-physical attacks can destroy equipment, force dimensional product changes, alter a product's mechanical characteristics, or endanger human lives. The manufacturing industry often relies on modern Quality Control (QC) tools to protect against quality losses, such as those that can occur from an attack. However, cyber-physical attacks can still be designed to avoid detection by traditional QC methods, which suggests a strong need for new and more robust QC tools. Such new tools should be able to prevent, or at least minimize, the effects of cyber-physical attacks on production systems. Unfortunately, little to no research has been done on using QC tools for cyber-physical security of production systems. Hence, the overarching goal of this work is to allow QC systems to be designed and used effectively as a second line of defense, when traditional cyber-security techniques fail and the production system is already breached. To this end, this work focuses on: 1) understanding the role of QC systems in cyber-physical attacks within manufacturing through developing a taxonomy encompassing the different layers involved; 2) identifying existing weaknesses in QC tools and exploring the effects of exploiting them by cyber-physical attacks; and 3) proposing more effective QC tools that can overcome existing weaknesses by introducing randomness to the tools, for better security against cyber-physical attacks in manufacturing. / Ph. D. / The recent technological developments in computers and networking have made systems, such as production systems, more vulnerable to attacks having both cyber and physical components; i.e., to cyber-physical attacks. In manufacturing, such attacks are not only capable of stealing valuable information, but can also destroy equipment, force physical product changes, alter product’s mechanical characteristics, or endanger human lives. Typically, the manufacturing industry have relied on various Quality Control (QC) tools, such as product inspection, to detect the effects caused by these attacks. However, these attacks could be still designed in a way to avoid detection by traditional QC methods, which suggests a need for new and more effective QC tools. Such new tools should be able to prevent, or at least minimize, the effects of these attacks in manufacturing. Unfortunately, almost no research has been done on using QC tools for securing production systems against these malicious attacks. Hence, the overarching goal of this work is to allow QC systems to be designed in a more effective manner to act as a second line of defense, when traditional cyber-security measures and attackers have already accessed the production system. To this end, this work focuses on: 1) understanding the role of QC systems during the attack; 2) identifying existing weaknesses in QC tools and determining the effects of exploiting them by the attack; and 3) proposing more effective QC tools, for better protection against these types of cyber-physical attacks in manufacturing.

This thesis proposes a new method for detecting malicious cyber-physical attacks on additive manufacturing (AM) systems. The method makes use of a physical hash, which links digital data to the manufactured part via a disconnected side-channel measurement system. The disconnection ensures that if the network and/or AM system become compromised, the manufacturer can still rely on the measurement system for attack detection. The physical hash takes the form of a QR code that contains a hash string of the nominal process parameters and toolpath. It is manufactured alongside the original geometry for the measurement system to scan and compare to the readings from its sensor suite. By taking measurements in situ, the measurement system can detect in real-time if the part being manufactured matches the designer's specification. A proof-of-concept validation was realized on a material extrusion machine. The implementation was successful and demonstrated the ability of this method to detect the existence (and absence) of malicious attacks on both process parameters and the toolpath. A case study for detecting changes to the toolpath is also presented, which uses a simple measurement of how long each layer takes to build. Given benchmark readings from a 30x30 mm square layer created on a material extrusion system, several modifications were able to be detected. The machine's repeatability and measurement technique's accuracy resulted in the detection of a 1 mm internal void, a 2 mm scaling attack, and a 1 mm skewing attack. Additionally, for a short to moderate length build of an impeller model, it was possible to detect a 0.25 mm change in the fin base thickness. A second case study is also presented wherein dogbone tensile test coupons were manufactured on a material extrusion system at different extrusion temperatures. This process parameter is an example of a setting that can be maliciously modified and have an effect on the final part strength without the operator's knowledge. The performance characteristics (Young's modulus and maximum stress) were determined to be statistically different at different extrusion temperatures (235 and 270 °C). / Master of Science / Additive Manufacturing (AM, also known as 3D printing) machines are cyber-physical systems and are therefore vulnerable to malicious attacks that can cause physical damage to the parts being manufactured or even to the machine itself. This thesis proposes a new method for detecting that an AM system has been hacked. Attacks are identified via a series of measurements taken by a measurement system that is disconnected from the main network. The disconnection ensures that if the network and/or AM system are hacked, the manufacturer can still rely on the measurement system for attack detection. The proposed method uses a physical hash to transfer information to the disconnected measurement system. This physical hash takes the form of a QR code and stores in it the nominal process parameters and toolpath of the build. It is manufactured alongside the original geometry for the measurement system to scan and compare to the readings from its sensor suite. By taking measurements in real-time, the measurement system can detect if the part being manufactured matches the designer’s specification. A proof-of-concept of the proposed method was realized on a common AM system. The implementation was successful and demonstrated the ability of this method to detect the existence of a malicious attack. A case study for detecting changes to the toolpath is also proposed using the simple measurement of how long each layer takes to build. Given benchmark readings of a part manufactured on the same technology as the proof-of-concept implementation, several modifications were able to be detected. The attack types tested were the insertion of an internal void, scaling the part, and skewing the part. A second case study is also presented where components were manufactured at different extrusion temperatures. By measuring the force required to break the parts, it was determined that temperature has an effect on the final part strength. This confirmed that malicious attacks targeting extrusion temperature are a plausible threat, and that the parameter should be measured in the proposed system.

The smart grid is the next-generation electrical infrastructure utilizing Information and Communication Technologies (ICTs), whose architecture is evolving from a utility-centric structure to a distributed Cyber-Physical System (CPS) integrated with a large-scale of renewable energy resources. However, meeting reliability objectives in the smart grid becomes increasingly challenging owing to the high penetration of renewable resources and changing weather conditions. Moreover, the cyber-physical attack targeted at the smart grid has become a major threat because millions of electronic devices interconnected via communication networks expose unprecedented vulnerabilities, thereby increasing the potential attack surface. This dissertation is aimed at developing novel game-theoretic and machine-learning techniques for addressing the reliability and security issues residing at multiple layers of the smart grid, including power distribution system reliability forecasting, risk assessment of cyber-physical attacks targeted at the grid, and cyber attack detection in the Advanced Metering Infrastructure (AMI) and renewable resources. This dissertation first comprehensively investigates the combined effect of various weather parameters on the reliability performance of the smart grid, and proposes a multilayer perceptron (MLP)-based framework to forecast the daily number of power interruptions in the distribution system using time series of common weather data. Regarding evaluating the risk of cyber-physical attacks faced by the smart grid, a stochastic budget allocation game is proposed to analyze the strategic interactions between a malicious attacker and the grid defender. A reinforcement learning algorithm is developed to enable the two players to reach a game equilibrium, where the optimal budget allocation strategies of the two players, in terms of attacking/protecting the critical elements of the grid, can be obtained. In addition, the risk of the cyber-physical attack can be derived based on the successful attack probability to various grid elements. Furthermore, this dissertation develops a multimodal data-driven framework for the cyber attack detection in the power distribution system integrated with renewable resources. This approach introduces the spare feature learning into an ensemble classifier for improving the detection efficiency, and implements the spatiotemporal correlation analysis for differentiating the attacked renewable energy measurements from fault scenarios. Numerical results based on the IEEE 34-bus system show that the proposed framework achieves the most accurate detection of cyber attacks reported in the literature. To address the electricity theft in the AMI, a Distributed Intelligent Framework for Electricity Theft Detection (DIFETD) is proposed, which is equipped with Benford’s analysis for initial diagnostics on large smart meter data. A Stackelberg game between utility and multiple electricity thieves is then formulated to model the electricity theft actions. Finally, a Likelihood Ratio Test (LRT) is utilized to detect potentially fraudulent meters.

Nous abordons des problèmes de sécurité dans des systèmes cyber-physiques industriels. Les attaques contre ces systèmes doivent être traitées à la fois en matière de sûreté et de sécurité. Les technologies de contrôles imposés par les normes industrielles, couvrent déjà la sûreté. Cependant, du point de vue de la sécurité, la littérature a prouvé que l’utilisation de techniques cyber pour traiter la sécurité de ces systèmes n’est pas suffisante, car les actions physiques malveillantes seront ignorées. Pour cette raison, on a besoin de mécanismes pour protéger les deux couches à la fois. Certains auteurs ont traité des attaques de rejeu et d’intégrité en utilisant une attestation physique, p. ex., le tatouage des paramètres physiques du système. Néanmoins, ces détecteurs fonctionnent correctement uniquement si les adversaires n’ont pas assez de connaissances pour tromper les deux couches. Cette thèse porte sur les limites mentionnées ci-dessus. Nous commençons en testant l’efficacité d’un détecteur qui utilise une signature stationnaire afin de détecter des actions malveillantes. Nous montrons que ce détecteur est incapable d’identifier les adversaires cyber-physiques qui tentent de connaître la dynamique du système. Nous analysons son ratio de détection sous la présence de nouveaux adversaires capables de déduire la dynamique du système. Nous revisitons le design original, en utilisant une signature non stationnaire, afin de gérer les adversaires visant à échapper à la détection. Nous proposons également une nouvelle approche qui combine des stratégies de contrôle et de communication. Toutes les solutions son validées à l’aide de simulations et maquettes d’entraînement / We address security issues in cyber-physical industrial systems. Attacks against these systems shall be handled both in terms of safety and security. Control technologies imposed by industrial standards already cover the safety dimension. From a security standpoint, the literature has shown that using only cyber information to handle the security of cyber-physical systems is not enough, since physical malicious actions are ignored. For this reason, cyber-physical systems have to be protected from threats to their cyber and physical layers. Some authors handle the attacks by using physical attestations of the underlying processes, f.i., physical watermarking to ensure the truthfulness of the process. However, these detectors work properly only if the adversaries do not have enough knowledge to mislead crosslayer data. This thesis focuses on the aforementioned limitations. It starts by testing the effectiveness of a stationary watermark-based fault detector, to detect, as well, malicious actions produced by adversaries. We show that the stationary watermark-based detector is unable to identify cyber-physical adversaries. We show that the approach only detects adversaries that do not attempt to get any knowledge about the system dynamics. We analyze the detection performance of the original design under the presence of adversaries that infer the system dynamics to evade detection. We revisit the original design, using a non-stationary watermark-based design, to handle those adversaries. We also propose a novel approach that combines control and communication strategies. We validate our solutions using numeric simulations and training cyber-physical testbeds