While significant research effort has been dedicated to wireless position location over the past decades, most location security aspects have been overlooked. Recently, with the proliferation of diverse wireless devices and the desire to determine their position, there is an increasing concern about the security of location information which can be spoofed or disrupted by adversaries or unreliable signal sources. This dissertation addresses the problem of securing a radio location system against location spoofing, specifically the characterization, analysis, detection, and localization of location spoofing attacks by focusing on fundamental location estimation issues. The objective of this dissertation is four-fold. First, it provides an overview of fundamental security issues for position location, particularly associated with range-based localization. Of particular interest are security risks and vulnerabilities in location estimation, types of localization attacks, and their impact. The second objective is to characterize the effects of signal strength and beamforming attacks on range estimates and the resulting position estimate. The characterization can be generalized to a variety of location spoofing attacks and provides insight into the anomalous behavior of range and location estimators when under attack. Through this effort we can also identify effective attacks that are of particular interest to attack detection and localization. The third objective is to develop an effective technique for attack detection which requires neither prior environmental nor statistical knowledge. This is accomplished by exploiting the bilateral behavior of a hybrid framework using two received signal strength (RSS) based location estimators. We show that the resulting approach is effective at detecting attacks with the detection rate increasing with the severity of the induced location error. The last objective of this dissertation is to develop a localization method resilient to attacks and other adverse effects. Since the detection and localization approach relies solely on RSS measurements in order to be applicable to a wide range of wireless systems and scenarios, this dissertation focuses on RSS-based position location. Nevertheless, many of the basic concepts and results can be applied to any range-based positioning system. / Ph. D.

The use of Machine Learning for malware detection is essential to counter the massive growth in malware types compared with the traditional signature-based detection system. However, machine learning models could also be extremely vulnerable and sensible to transferable adversarial example (AE) attacks. The transfer AE attack does not require extra information from the victim model such as gradient information. Researchers explore mainly 2 lines of transfer-based adversarial example attacks: ensemble models and ensemble samples. \\ Although comprehensive innovations and progress have been achieved in transfer AE attacks, few works have investigated how these techniques perform in malware data. Besides, generating adversarial examples on an android APK file is not as easy and convenient as it is on image data since the generated AE of malware should also remain its functionality and executability after perturbation. Therefore, it is urgent to validate whether previous methodologies could still have their effect on malware considering the differences compared to image data. \\ In this thesis, we first have a thorough literature review for the AE attacks on malware data and general transfer AE attacks. Then we design our algorithm for the transfer AE attack. We formulate the optimization problem based on the intuition that the contribution evenness of features towards the final prediction result is highly correlated to the AE transferability. We then solve the optimization problem by gradient descent and evaluate it through extensive experiments. Implementing and experimenting with the state-of-the-art AE algorithms and transferability enhancement techniques, we analyze and summarize the weaknesses and strengths of each method. / Master of Science / Machine learning models have been widely applied to malware detection systems in recent years due to the massive growth in malware types. However, these models are vulnerable to adversarial attacks. Malicious attackers can add some small imperceptible perturbations to the original testing samples and mislead the classification results at a very low cost. Research on adversarial attacks would help us gain a better understanding of the attacker's side and inspire defenses against them. Among all adversarial attacks, the transfer-based adversarial example attack is one of the most devastating attacks since it does not require extra information from the targeted victim model such as gradient information or query from the model. Although plenty of researchers has explored the transfer AE attack lately, few works focus on malware (e.g., Android) data. Compared with image data, perturbing malware is more complicated and challenging since the generated adversarial examples of malware need to remain functionality and executability. To validate how transfer AE attack methods perform on malware, we implement the state-of-the-art (SOTA) works in this thesis and experiment with them on real Android data. Besides, we develop a new transfer-based AE attack method based on the contribution of each feature for generating AE. We then do comprehensive evaluations and draw comparisons between SOTA works and our proposed method.

Mission impact assessment (MIA) research has been explored to evaluate the performance and effectiveness of a mission system, such as enterprise networks with organizational missions and military or tactical mission teams with assigned missions. The key components in such mission systems, including assets, services, tasks, vulnerability, attacks, and defenses, are interdependent, and their impacts are interwoven. However, the current state-of-the-art MIA approaches have less studied such interdependencies. In addition, they have not modeled strategic attack-defense interactions under partial observability. In this work, we propose a novel MIA framework that assesses measures of performance (MoP) or measures of effectiveness (MoE) based on the service requirements (e.g., correctness or timeliness) of a given mission system based on full and comprehensive modeling and simulation of the key system components and their interdependencies. Particularly, we model intelligent attack-defense strategy selections based on hypergame theory, which allows considering uncertainty in estimating each player's hypergame expected utility (HEU) for its best strategy selection. As the case study, we consider an Internet-of-Things (IoT)-based mission system aiming to accurately and timely detect an object, given stringent accuracy and time constraints for successful mission completion. Via extensive simulation experiments, we validate the quality of the proposed MIA tool in its inference accuracy of the mission performance under a wide range of different environmental settings hindering the mission performance assessment and attack-defense interactions. Our results prove that the developed MIA framework shows a sufficiently high inference accuracy (e.g., 80%) even with a small portion of the training dataset (e.g., 20-50%). We also found the MIA can better assess the system's mission performance when attackers exhibit clearer patterns to take more strategic actions using hypergame theory. / Master of Science / In our increasingly interconnected world, mission systems play a crucial role, whether in organizational networks or tactical military operations. We often evaluate these systems to ensure they perform effectively, but there's more to it than meets the eye. Imagine an intricate web of resources, tasks, services, assaults, and defenses that are intertwined and have an impact on one another. The strategic interactions of attack and defense in uncertain environments have been majorly ignored by conventional techniques for mission impact assessment (MIA). Our research introduces a new way of thinking about MIA. We've developed a framework that delves deep into the heart of mission systems, considering how each component affects the others. This comprehensive approach considers not just what's happening but also the interplay of actions and reactions. Hypergame theory, a technique that enables us to model intelligent choices in the face of uncertainty, is at the foundation of our approach. Imagine it as a chess game in which players must predict their opponents' moves and adjust their strategies appropriately. In our case study, we used an Internet-of-Things (IoT)-based mission system tasked with timely and accurate object detection to apply this architecture. In this mission system, both cyber attackers, whose aim is to compromise the mission, and cyber defenders, whose aim is to ensure mission success, are present, and they use the proposed hypergame-based decision-making to perform intelligent actions. What did we find? Through extensive simulations, we confirmed the effectiveness of our MIA framework. Even with limited training data, our tool demonstrated a remarkable 80% accuracy in assessing mission performance. Moreover, it excelled when attackers followed discernible patterns, allowing us to predict and respond strategically. In simpler terms, our research provides a valuable tool for evaluating the success of mission systems in our increasingly connected world. It goes beyond surface-level assessments, considering the intricate relationships between system components and the dynamic nature of strategic decision-making. Ultimately, our framework empowers us to ensure mission success in an ever-evolving landscape.

The primary focus of this doctoral dissertation is to investigate the safety and robustness of deep models. Our objective is to thoroughly analyze and introduce innovative methodologies for cultivating robust representations under diverse circumstances. Deep neural networks (DNNs) have emerged as fundamental components in recent advancements across various tasks, including image recognition, semantic segmentation, and object detection. Representation learning stands as a pivotal element in the efficacy of DNNs, involving the extraction of significant features from data through mechanisms like convolutional neural networks (CNNs) applied to image data. In real-world applications, ensuring the robustness of these features against various adversarial conditions is imperative, thus emphasizing robust representation learning. Through the acquisition of robust representations, DNNs can enhance their ability to generalize to new data, mitigate the impact of label noise and domain shifts, and bolster their resilience against external threats, such as backdoor attacks. Consequently, this dissertation explores the implications of robust representation learning in three principal areas: i) Backdoor Attack, ii) Backdoor Defense, and iii) Noisy Labels. First, we study the backdoor attack creation and detection from different perspectives. Backdoor attack addresses AI safety and robustness issues where an adversary can insert malicious behavior into a DNN by altering the training data. Second, we aim to remove the backdoor from DNN using two different types of defense techniques: i) training-time defense and ii) test-time defense. training-time defense prevents the model from learning the backdoor during model training whereas test-time defense tries to purify the backdoor model after the backdoor has already been inserted. Third, we explore the direction of noisy label learning (NLL) from two perspectives: a) offline NLL and b) online continual NLL. The representation learning under noisy labels gets severely impacted due to the memorization of those noisy labels, which leads to poor generalization. We perform uniform sampling and contrastive learning-based representation learning. We also test the algorithm efficiency in an online continual learning setup. Furthermore, we show the transfer and adaptation of learned representations in one domain to another domain, e.g. source free domain adaptation (SFDA). We study the impact of noisy labels under SFDA settings and propose a novel algorithm that produces state-of-the-art (SOTA) performance.

Many real word networks exhibit cascading phenomena, e.g., disease outbreaks in social contact networks, malware propagation in computer networks, failures in cyber-physical systems such as power grids. As they grow in size and complexity, their security becomes increasingly important. In this thesis, we address the problems of controlling cascading failures in various network settings. We address the cascading phenomena which are either natural (e.g., disease outbreaks) or malicious (e.g., cyber attacks). We consider the nodes of a network as being individually or collectively controlled by self-interested autonomous agents and study their strategic decisions in the presence of these failure cascades. There are many models of cascading failures which specify how a node would fail when some neighbors have failed, such as: (i) epidemic spread models in which the cascading can be viewed as a natural and stochastic process and (ii) cyber attack models where the cascade is driven by malicious intents. We present our analyses and algorithms for these models in two parts. Part I focuses on problems of controlling epidemic spread. Epidemic outbreaks are generally modeled as stochastic diffusion processes. In particular, we consider the SIS model on networks. There exist heuristic centralized approaches in the literature for containing epidemic spread in SIS/SIR models; however no rigorous performance bounds are known for these approaches. We develop algorithms with provable approximation guarantees that involve either protective intervention (e.g., vaccination) or link removal (e.g., unfriending). Our approach relies on the characterization of the SIS model in terms of the spectral radius of the network. The centralized approaches, however, are sometimes not feasible in practice. For example, targeted vaccination is often not feasible because of limited compliance to directives. This issue has been addressed in the literature by formulating game theoretic models for the containment of epidemic spread. However they generally assume simplistic propagation models or homogeneous network structures. We develop novel game formulations which rely on the spectral characterization of the SIS model. In these formulations, the failures start from a random set of nodes and propagate through the network links. Each node acts as a self-interested agent and makes strategic intervention decisions (e.g., taking vaccination). Each agent decides its strategy to optimize its payoff (modeled by some payoff function). We analyze the complexity of finding Nash equilibria (NE) and study the structure of NE for different networks in these game settings. Part II focuses on malware spread in networks. In cybersecurity literature malware spreads are often studied in the framework of ``attack graph" models. In these models, a node represents either a physical computing unit or a network configuration and an edge represents a physical or logical vulnerability dependency. A node gets compromised if a certain set of its neighbors are compromised. Attack graphs describe explicit scenarios in which a single vulnerability exploitation cascades further into the network exploiting inherent dependencies among the network components. Attack graphs are used for studying cascading effects in many cybersecurity applications, e.g., component failure in enterprise networks, botnet spreads, advanced persistent attacks. One distinct feature of cyber attack cascades is the stealthy nature of the attack moves. Also, cyber attacks are generally repeated. How to control stealthy and repeated attack cascades is an interesting problem. Dijk et. al.~cite{van2013flipit} first proposed a game framework called ``FlipIt" for reasoning about the stealthy interaction between a defender and an attacker over the control of a system resource. However, in cybersecurity applications, systems generally consists of multiple resources connected by a network. Therefore it is imperative to study the stealthy attack and defense in networked systems. We develop a generalized framework called ``FlipNet" which extends the work of Dijk et. al.~cite{van2013flipit} for network. We present analyses and algorithms for different problems in this framework. On the other hand, if the security of a system is limited to the vulnerabilities and exploitations that are known to the security community, often the objective of the system owner is to take cost-effective steps to minimize potential damage in the network. This problem has been formulated in the cybersecurity literature as hardening attack graphs. Several heuristic approaches have been shown in the litrature so far but no algorithmic analysis have been shown. We analyze the inherent vulnerability of the network and present approximation hardening algorithms. / Ph. D.

As buildings evolve towards the envisioned smart building paradigm, smart buildings' cyber-security issues and physical security issues are mingling. Although research studies have been conducted to detect and prevent physical (or cyber) intrusions to smart building systems(SBS), it is still unknown (1) how one type of intrusion facilitates the other, and (2) how such synergic attacks compromise the security protection of whole systems. To investigate both research questions, the author proposes a graph-based testbed to simulate cyber-physical attacks on smart buildings. The testbed models both cyber and physical accesses of a smart building in an integrated graph, and simulates diverse cyber-physical attacks to assess their synergic impacts on the building and its systems. In this thesis, the author presents the testbed design and the developed prototype, SHSIM. An experiment is conducted to simulate attacks on multiple smart home designs and to demonstrate the functions and feasibility of the proposed simulation system. / Master of Science / A smart home/building is a residence containing multiple connected devices which enable remote monitoring, automation, and management of appliances and systems, such as lighting, heating, entertainment, etc. Since the early 2000s, this concept of a smart home has becomequite popular due to rapid technological improvement. However, it brings with it a lot of security issues. Typically, security issues related to smart homes can be classified into two types - (1) cybersecurity and (2) physical security. The cyberattack involves hacking into a network to gain remote access to a system. The physical attack deals with unauthorized access to spaces within a building by damaging or tampering with access control. So far the two kinds of attacks on smart homes have been studied independently. However, it is still unknown (1) how one type of attack facilitates the other, and (2) how the combination of two kinds of attacks compromises the security of the whole smart home system. Thus, to investigate both research questions, we propose a graph-based approach to simulate cyber-physical attacks on smart homes/buildings. During the process, we model the smart home layout into an integrated graph and apply various cyber-physical attacks to assess the security of the smart building. In this thesis, I present the design and implementation of our tool, SHSIM. Using SHSIM we perform various experiments to mimic attacks on multiple smart home designs. Our experiments suggest that some current smart home designs are vulnerable to cyber-physical attacks

Modern innovation is a driving force behind increased spectrum crowding. Several studies performed by the National Telecommunications and Information Administration (NTIA), Federal Communications Commission (FCC), and other groups have proposed Dynamic Spectrum Access (DSA) as a promising solution to alleviate spectrum crowding. The spectrum assignment decisions in DSA will be made by a centralized entity referred to as as spectrum access system (SAS); however, maintaining spectrum utilization information in SAS presents privacy risks, as sensitive Incumbent User (IU) operation parameters are required to be stored by SAS in order to perform spectrum assignments properly. These sensitive operation parameters may potentially be compromised if SAS is the target of a cyber attack or an inference attack executed by a secondary user (SU). In this thesis, we explore the operational security of IUs in SAS-based DSA systems and propose a novel privacy-preserving SAS-based DSA framework, Suspicion Zone SAS (SZ-SAS), the first such framework which protects against both the scenario of inference attacks in an area with sparsely distributed IUs and the scenario of untrusted or compromised SAS. We then define modifications to the SU inference attack algorithm, which demonstrate the necessity of applying obfuscation to SU query responses. Finally, we evaluate obfuscation schemes which are compatible with SZ-SAS, verifying the effectiveness of such schemes in preventing an SU inference attack. Our results show SZ-SAS is capable of utilizing compatible obfuscation schemes to prevent the SU inference attack, while operating using only homomorphically encrypted IU operation parameters. / Master of Science / Dynamic Spectrum Access (DSA) allows users to opportunistically access spectrum resources which were previously reserved for use by specified parties. This spectrum sharing protocol has been identified as a potential solution to the issue of spectrum crowding. This sharing will be accomplished through the use of a centralized server, known as a spectrum access system (SAS). However, current SAS-based DSA proposals require users to submit information such as location and transmission properties to SAS. The privacy of these users is of the utmost importance, as many existing users in these spectrum bands are military radars and other users for which operational security is pivotal. Storing the information for these users in a central database can be an major privacy issue, as this information could be leaked if SAS is compromised by a malicious party. Additionally, malicious secondary users (SUs) may perform an inference attack, which could also reveal the location of these military radars. In this thesis, we demonstrate a SAS-framework, SZ-SAS, which allows SAS to function without direct knowledge of user information. We also propose techniques for mitigating the inference attack which are compatible with SZ-SAS

In our daily life, we are increasingly putting our trust in embedded software applications, which run on a range of processor-based embedded systems from smartcards to pay-TV units. This trend expands the threat model of embedded applications from software into hardware. Over the last 20 years, fault attacks have emerged as an important class of hardware attacks against embedded software security. In fault attacks, an adversary breaks the security by injecting well chosen, targeted faults during the execution of embedded software, and systematically analyzing softwares fault response. In this work, we propose cycle-accurate and fully digital techniques that can efficiently detect different types of fault attacks. The detection methods are low-cost regarding the area and power consumption and can be easily implemented using the standard cell based VLSI design flow. In addition to the architecture of the detectors, we present a detailed analysis of the design considerations that affect the cost and accuracy of the detectors. The functionality of the detectors is validated by implementing on ASIC and FPGA platforms (Spartan-6, Cyclone IV). Additionally, the proposed detection methods have demonstrated to successfully detect all of the injected faults without any false alarm. / Master of Science / Embedded systems nowadays play a very crucial role in day to day life. They are always gathering sensitive and private data of the users. So they become an attractive target for the attackers to steal this important data. As a result, the security of these devices has become a grave concern. Fault attacks are a class of hardware attacks where the attacker injects faults into the system while it is executing a known program and observes the reaction. The abnormal reactions of the system are later analyzed to obtain the valuable data. Several mechanisms to detect such attacks exist in the literature, but they are not very effective. In this work, we first analyze the effect of different types of fault attacks on the embedded processor. Then we propose various low-cost digital techniques that can efficiently detect these attacks.

No / There is developing interest in the analytical use of human hair from archaeological contexts in key research areas such as DNA, trace elemental and isotopic analyses. Other human tissues, especially bone, that have been used for trace element, isotopic and DNA analyses have had extensive study concerning their diagenesis, but this has not been done for hair. Consideration must be given to the complex interaction of hair with its buried environment, thereby laying a firm basis for the use of hair in future research. Since human hair is known to survive under a diverse range of environmental conditions, a pilot study has investigated the basic processes of hair degradation, using samples from different climatic zones and burial types. Variation in the degree of preservation of archaeological hair was characterized by light microscopy, electron microscopy, and FT-Raman spectroscopy, relating morphological change of the surface and internal structure of hair to its biochemical integrity. The results demonstrate a breakdown of cortical cell boundaries and disruption of the cuticular layering, coupled with infiltration of material from the burial matrix that suggests a progressive loss of cohesion that is in part due to microbiological activity. Medullated hair is shown to be more susceptible to physical breakdown by providing two routes for microbial and environmental attack. At the molecular level the proteinaceous component undergoes alteration, and the S-S cystine linkages, responsible for the strength and resilience of hair in living individuals, are lost.

With the increase of requesting personal information online, unauthorized disclosure of user privacy is a significant problem faced by today's Internet. As a typical identity theft, phishing usually employs fraudulent emails and spoofed web sites to trick unsuspecting users into divulging their private information. Even legitimate web sites may collect private information from unsophisticated users such as children for commercial purposes without their parents' consent. The Children's Online Privacy Protection Act (COPPA) of 1998 was enacted in reaction to the widespread collection of information from children and subsequent abuses identified by the Federal Trade Commission (FTC). COPPA is aimed at protecting child's privacy by requiring parental consent before collecting information from children under thirteen. In this thesis, we propose two solutions for ensuring user online privacy. By analyzing common characteristics of phishing pages, we propose a client-side tool, Trident, which works as a browser plug-in for filtering phishes. The experiment results show that Trident can identify 98-99% online and valid phishing pages, as well as automatically validate legitimate pages. To protect child's privacy, we introduce the POCKET (parental online consent on kids' electronic privacy) framework, which is a technically feasible and legally sound solution to enforce COPPA. Parents answer a questionnaire on their privacy requirements and the POCKET user agent generates a privacy preferences file. Meantime, the merchants are required to possess a privacy policy that is authenticated by a trusted third party. Only web sites that possess and adhere to their privacy policies are allowed to collect child's information; web sites whose policies do not match the client's preferences are blocked. POCKET framework incorporates a transaction protocol to secure the data exchange between an authenticated client and a POCKET-compliant merchant. / Master of Science