abstract: The field of cyber-defenses has played catch-up in the cat-and-mouse game of finding vulnerabilities followed by the invention of patches to defend against them. With the complexity and scale of modern-day software, it is difficult to ensure that all known vulnerabilities are patched; moreover, the attacker, with reconnaissance on their side, will eventually discover and leverage them. To take away the attacker's inherent advantage of reconnaissance, researchers have proposed the notion of proactive defenses such as Moving Target Defense (MTD) in cyber-security. In this thesis, I make three key contributions that help to improve the effectiveness of MTD. First, I argue that naive movement strategies for MTD systems, designed based on intuition, are detrimental to both security and performance. To answer the question of how to move, I (1) model MTD as a leader-follower game and formally characterize the notion of optimal movement strategies, (2) leverage expert-curated public data and formal representation methods used in cyber-security to obtain parameters of the game, and (3) propose optimization methods to infer strategies at Strong Stackelberg Equilibrium, addressing issues pertaining to scalability and switching costs. Second, when one cannot readily obtain the parameters of the game-theoretic model but can interact with a system, I propose a novel multi-agent reinforcement learning approach that finds the optimal movement strategy. Third, I investigate the novel use of MTD in three domains-- cyber-deception, machine learning, and critical infrastructure networks. I show that the question of what to move poses non-trivial challenges in these domains. To address them, I propose methods for patch-set selection in the deployment of honey-patches, characterize the notion of differential immunity in deep neural networks, and develop optimization problems that guarantee differential immunity for dynamic sensor placement in power-networks. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2020

As the use of low-power and low-resource embedded devices continues to increase dramatically with the introduction of new Internet of Things (IoT) devices, security techniques are necessary which are compatible with these devices. This research advances the knowledge in the area of cybersecurity for the IoT through the exploration of a moving target defense to apply for limiting the time attackers may conduct reconnaissance on embedded systems while considering the challenges presented from IoT devices such as resource and performance constraints. We introduce the design and optimizations for µMT6D, a Micro-Moving Target IPv6 Defense, including a description of the modes of operation and use of lightweight hash algorithms. Through simulations and experiments µMT6D is shown to be viable for use on low power and low resource embedded devices in terms of footprint, power consumption, and energy consumption increases in comparison to the given security benefits. Finally, this provides information on other future considerations and possible avenues of further experimentation and research. / Doctor of Philosophy / This research aims to advance knowledge in the area of cybersecurity for the Internet of Things through the exploration and validation of a moving target defense to apply for limiting the time attackers may conduct reconnaissance on low powered embedded system devices considering the challenges presented from IoT devices such as resource and performance constraints. When an attack is carried out against a network, reconnaissance is utilized to identify the target machine or device. Limiting the time for reconnaissance, therefore has a direct impact on the ability of an adversary to carry out an attack. Many of the security techniques utilized today do not fit the IoT constraints. Research in this area is just beginning and security is often not considered. Sensors collecting and sending information can be compromised both through the network and access to the physical devices. How can these devices securely send information? How can these devices withstand attacks aiming to stop their functionality or to gain information? There are many aspects which need to be investigated to understand security vulnerabilities and potential defenses. As our technologies evolve our security defenses need to evolve as well. My research aims to further the understanding of the security of the IoT devices which have quickly become pervasive in our society. This research will expand the knowledge of the ability to safe guard connected devices from cyber-attacks and provide insight into the space and performance requirements of a technique previously only used on large scale systems. By designing, implementing experimental prototypes, and conducting simulations and experiments this research assesses the viable use of a Micro Moving Target IPv6 Defense (µMT6D).

According to security.org, as of 2023, 65% of credit card (CC) users in the US have been subjected to fraud at some point in their lives, which equates to about 151 million Americans. The proliferation of advanced machine learning (ML) algorithms has also contributed to detecting credit card fraud (CCF). However, using a single or static ML-based defense model against a constantly evolving adversary takes its structural advantage, which enables the adversary to reverse engineer the defense's strategy over the rounds of an iterated game. This paper proposes an adaptive moving target defense (MTD) approach based on deep reinforcement learning (DRL), termed RESONANT to identify the optimal switching points to another ML classifier for credit card fraud detection. It identifies optimal moments to strategically switch between different ML-based defense models (i.e., classifiers) to invalidate any adversarial progress and always stay a step ahead of the adversary. We take this approach in an iterated game theoretic manner where the adversary and defender take turns to take their action in the CCF detection contexts. Via extensive simulation experiments, we investigate the performance of our proposed RESONANT against that of the existing state-of-the-art counterparts in terms of the mean and variance of detection accuracy and attack success ratio to measure the defensive performance. Our results demonstrate the superiority of RESONANT over other counterparts, including static and naïve ML and MTD selecting a defense model at random (i.e., Random-MTD). Via extensive simulation experiments, our results show that our proposed RESONANT can outperform the existing counterparts up to two times better performance in detection accuracy using AUC (i.e., Area Under the Curve of the Receiver Operating Characteristic (ROC) curve) and system security against attacks using attack success ratio (ASR). / Master of Science / According to security.org, as of 2023, 65% of credit card (CC) users in the US have been subjected to fraud at some point in their lives, which equates to about 151 million Americans. The proliferation of advanced machine learning (ML) algorithms has also contributed to detecting credit card fraud (CCF). However, using a single or static ML-based defense model against a constantly evolving adversary takes its structural advantage, which enables the adversary to reverse engineer the defense's strategy over the rounds of an iterated game. This paper proposes an adaptive defense approach based on artificial intelligence (AI), termed RESONANT, to identify the optimal switching points to another ML classifiers for credit card fraud detection. It identifies optimal moments to strategically switch between different ML-based defense models (i.e., classifiers) to invalidate any adversarial progress and always stay a step ahead of the adversary. We take this approach in an iterated game theoretic manner where the adversary and defender take turns to take their action in the CCF detection contexts. Via extensive simulation experiments, we investigate the performance of our proposed RESONANT against that of the existing state-of-the-art counterparts in terms of the mean and variance of detection accuracy and attack success ratio to measure the defensive performance. Our results demonstrate the superiority of RESONANT over other counterparts, showing that our proposed RESONANT can outperform the existing counterparts by up to two times better performance in detection accuracy and system security against attacks.

Cyber-physical systems (CPSs) are systems that integrate physical elements with a cyber layer that enables sensing, monitoring, and processing the data from the physical components. Examples of CPSs include autonomous vehicles, unmanned aerial vehicles (UAVs), smart grids, and the Internet of Things (IoT). In particular, many critical infrastructure (CI) that are vital to our modern day cities and communities, are CPSs. This wide range of CPSs domains represents a cornerstone of smart cities in which various CPSs are connected to provide efficient services. However, this level of connectivity has brought forward new security challenges and has left CPSs vulnerable to many cyber-physical attacks and disruptive events that can utilize the cyber layer to cause damage to both cyber and physical components. Addressing these security and operation challenges requires developing new security solutions to prevent and mitigate the effects of cyber and physical attacks as well as improving the CPSs response in face of disruptive events, which is known as the CPS resilience. To this end, the primary goal of this dissertation is to develop novel analytical tools that can be used to study, analyze, and optimize the resilience and security of critical CPSs. In particular, this dissertation presents a number of key contributions that pertain to the security and the resilience of multiple CPSs that include power systems, the Internet of Things (IoT), UAVs, and transportation networks. First, a mathematical framework is proposed to analyze and mitigate the effects of GPS spoofing attacks against UAVs. The proposed framework uses system dynamics to model the optimal routes which UAVs can follow in normal operations and under GPS spoofing attacks. A countermeasure mechanism, built on the premise of cooperative localization, is then developed to mitigate the effects of these GPS spoofing attacks. To practically deploy the proposed defense mechanism, a dynamic Stackelberg game is formulated to model the interactions between a GPS spoofer and a drone operator. The equilibrium strategies of the game are analytically characterized and studied through a novel, computationally efficient algorithm. Simulation results show that, when combined with the Stackelberg strategies, the proposed defense mechanism will outperform baseline strategy selection techniques in terms of reducing the possibility of UAV capture. Next, a game-theoretic framework is developed to model a novel moving target defense (MTD) mechanism that enables CPSs to randomize their configurations to proactive deter impending attacks. By adopting an MTD approach, a CPS can enhance its security against potential attacks by increasing the uncertainty on the attacker. The equilibrium of the developed single-controller, stochastic MTD game is then analyzed. Simulation results show that the proposed framework can significantly improve the overall utility of the defender. Third, the concept of MTD is coupled with new cryptographic algorithms for enhancing the security of an mHealth Internet of Things (IoT) system. In particular, using a combination of theory and implementation, a framework is introduced to enable the IoT devices to update their cryptographic keys locally to eliminate the risk of being revealed while they are shared. Considering the resilience of CPSs, a novel framework for analyzing the component- and system-level resilience of CIs is proposed. This framework brings together new ideas from Bayesian networks and contract theory – a Nobel prize winning theory – to define a concrete system-level resilience index for CIs and to optimize the allocation of resources, such as redundant components, monitoring devices, or UAVs to help those CIs improve their resilience. In particular, the developed resilience index is able to account for the effect of CI components on the its probability of failure. Meanwhile, using contract theory, a comprehensive resource allocation framework is proposed enabling the system operator to optimally allocate resources to each individual CI based on its economic contribution to the entire system. Simulation results show that the system operator can economically benefit from allocating the resources while dams can have a significant improvement in their resilience indices. Subsequently, the developed contract-theoretic framework is extended to account for cases of asymmetric information in which the system operator has only partial information about the CIs being in some vulnerability and criticality levels. Under such asymmetry, it is shown that the proposed approach maximizes the system operator's utility while ensuring that no CI has an incentive to ask for another contract. Next, a proof-of-concept framework is introduced to analyze and improve the resilience of transportation networks against flooding. The effect of flooding on road capacities and on the free-flow travel time, is considered for different rain intensities and roads preparedness. Meanwhile, the total system's travel time before and after flooding is evaluated using the concept of a Wardrop equilibrium. To this end, a proactive mechanism is developed to reduce the system's travel time, after flooding, by shifting capacities (available lanes) between same road sides. In a nutshell, this dissertation provides a suite of analytical techniques that allow the optimization of security and resilience across multiple CPSs. / Doctor of Philosophy / Cyber-physical systems (CPSs) have recently been used in many application domains because of their ability to integrate physical elements with a cyber layer allowing for sensing, monitoring, and remote controlling. This pervasive use of CPSs in different applications has brought forward new security challenges and threats. Malicious attacks can now leverage the connectivity of the cyber layer to launch remote attacks and cause damage to the physical components. Taking these threats into consideration, it became imperative to ensure the security of CPSs. Given that many CPSs provide critical services, for instance many critical infrastructure (CI) are CPSs such as smart girds and nuclear reactors; it is then inevitable to ensure that these critical CPSs can maintain proper operation. One key measure of the CPS’s functionality, is resilience which evaluates the ability of a CPS to deliver its designated service under potentially disruptive situations. In general, resilience measures a CPS’s ability to adapt or rapidly recover from disruptive events. Therefore, it is crucial for CPSs to be resilient in face of potential failures. To this end, the central goal of this dissertation is to develop novel analytical frameworks that can evaluate and improve security and resilience of CPSs. In these frameworks, cross-disciplinary tools are used from game theory, contract theory, and optimization to develop robust analytical solutions for security and resilience problems. In particular, these frameworks led to the following key contributions in cyber security: developing an analytical framework to mitigate the effects of GPS spoofing attacks against UAVs, introducing a game-theoretic moving target defense (MTD) framework to improve the cyber security, and securing data privacy in m-health Internet of Things (IoT) networks using a MTD cryptographic framework. In addition, the dissertation led to the following contributions in CI resilience: developing a general framework using Bayesian Networks to evaluate and improve the resilience of CIs against their components failure, introducing a contract-theoretic model to allocate resources to multiple connected CIs under complete and asymmetric information scenarios, providing a proactive plan to improve the resilience of transportation networks against flooding, and, finally, developing an environment-aware framework to deploy UAVs in disaster-areas.

Traditional static network addressing allows attackers the incredible advantage of taking time to plan and execute attacks against a network. To counter, Moving Target IPv6 Defense (MT6D) provides a network host obfuscation technique that dynamically obscures network and transport layer addresses. Software driven implementations have posed many challenges, namely, constant code maintenance to remain compliant with all library and kernel dependencies, less than optimal throughput, and the requirement for a dedicated general purpose hardware. The work of this thesis presents Network Security Processor and Hardware Engine for MT6D (HE-MT6D) to overcome these challenges. HE-MT6D is a soft core Intellectual Property (IP) block developed in full Register Transfer Level (RTL) and is the first hardware-oriented design of MT6D. Major contributions of HE-MT6D include the complete separation of the data and control planes, development of a nonlinear Complex Instruction Set Computer (CISC) Network Security Processor for in-flight packet modification, a specialized Packet Assembly language, a configurable and a parallelized memory search through tag-based Hybrid Content Addressable Memory (HCAM) L1 write-through cache, full RTL Network Time Protocol version 4 hardware module, and a modular crypto engine. HE-MT6D supports multiple nodes and provides 1,025% throughput performance increase over earlier C-based MT6D at 863 Mbps with full encapsulation and decapsulation, and it matches bare wire throughput performance for all other traffic. The HE-MT6D IP block can be configured as an independent physical gateway device, built as embedded Application Specific Integrated Circuit (ASIC), or serve as a System on Chip (SoC) integrated submodule. / Master of Science

abstract: The use of reactive security mechanisms in enterprise networks can, at times, provide an asymmetric advantage to the attacker. Similarly, the use of a proactive security mechanism like Moving Target Defense (MTD), if performed without analyzing the effects of security countermeasures, can lead to security policy and service level agreement violations. In this thesis, I explore the research questions 1) how to model attacker-defender interactions for multi-stage attacks? 2) how to efficiently deploy proactive (MTD) security countermeasures in a software-defined environment for single and multi-stage attacks? 3) how to verify the effects of security and management policies on the network and take corrective actions? I propose a Software-defined Situation-aware Cloud Security framework, that, 1) analyzes the attacker-defender interactions using an Software-defined Networking (SDN) based scalable attack graph. This research investigates Advanced Persistent Threat (APT) attacks using a scalable attack graph. The framework utilizes a parallel graph partitioning algorithm to generate an attack graph quickly and efficiently. 2) models single-stage and multi-stage attacks (APTs) using the game-theoretic model and provides SDN-based MTD countermeasures. I propose a Markov Game for modeling multi-stage attacks. 3) introduces a multi-stage policy conflict checking framework at the SDN network's application plane. I present INTPOL, a new intent-driven security policy enforcement solution. INTPOL provides a unified language and INTPOL grammar that abstracts the network administrator from the underlying network controller's lexical rules. INTPOL develops a bounded formal model for network service compliance checking, which significantly reduces the number of countermeasures that needs to be deployed. Once the application-layer policy conflicts are resolved, I utilize an Object-Oriented Policy Conflict checking (OOPC) framework that identifies and resolves rule-order dependencies and conflicts between security policies. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2020