No description available.

One goal of systems biology is to predict and modify the behavior of biological networks by accurately monitoring and modeling their responses to certain types of perturbations. The construction of mathematical models based on observation of these responses, referred to as reverse engineering, is an important step in elucidating the structure and dynamics of such networks. Continuous models, described by systems of differential equations, have been used to reverse engineer biochemical networks. Of increasing interest is the use of discrete models, which may provide a conceptual description of the network. In this dissertation we introduce a discrete modeling approach, rooted in computational algebra, to reverse-engineer networks from experimental time series data. The algebraic method uses algorithmic tools, including Groebner-basis techniques, to build the set of all discrete models that fit time series data and to select minimal models from this set. The models used in this work are discrete-time finite dynamical systems, which, when defined over a finite field, are described by systems of polynomial functions. We present novel reverse-engineering algorithms for discrete models, where each algorithm is suitable for different amounts and types of data. We demonstrate the effectiveness of the algorithms on simulated networks and conclude with a description of an ongoing project to reverse-engineer a real gene regulatory network in yeast. / Ph. D.

This dissertation develops algebraic theory for discrete models in systems biology. Many discrete model types can be translated into the framework of polynomial dynamical systems (PDS), that is, time- and state-discrete dynamical systems over a finite field where the transition function for each variable is given as a polynomial. This allows for using a range of theoretical and computational tools from computer algebra, which results in a powerful computational engine for model construction, parameter estimation, and analysis methods. Formal definitions and theorems for PDS and the concept of PDS as models of biological systems are introduced in section 1.3. Constructing a model for given time-course data is a challenging problem. Several methods for reverse-engineering, the process of inferring a model solely based on experimental data, are described briefly in section 1.3. If the underlying dependencies of the model components are known in addition to experimental data, inferring a "good" model amounts to parameter estimation. Chapter 2 describes a parameter estimation algorithm that infers a special class of polynomials, so called nested canalyzing functions. Models consisting of nested canalyzing functions have been shown to exhibit desirable biological properties, namely robustness and stability. The algorithm is based on the parametrization of nested canalyzing functions. To demonstrate the feasibility of the method, it is applied to the cell-cycle network of budding yeast. Several discrete model types, such as Boolean networks, logical models, and bounded Petri nets, can be translated into the framework of PDS. Section 3 describes how to translate agent-based models into polynomial dynamical systems. Chapter 4, 5, and 6 are concerned with analysis of complex models. Section 4 proposes a new method to identify steady states and limit cycles. The method relies on the fact that attractors correspond to the solutions of a system of polynomials over a finite field, a long-studied problem in algebraic geometry which can be efficiently solved by computing Gröbner bases. Section 5 introduces a bit-wise implementation of a Gröbner basis algorithm for Boolean polynomials. This implementation has been incorporated into the core engine of Macaulay 2. Chapter 6 discusses bistability for Boolean models formulated as polynomial dynamical systems. / Ph. D.

Internet-of-Things (IoT) devices are increasing in popularity due to their ability to help automate many aspects of daily life while performing these necessary duties on billions of low-power appliances. However, the perks of these small devices also come with additional constraints to security. Security always has been an issue with the rise of cryptographic backdoors and hackers reverse engineering the security protocols within devices to reveal the original state that was encrypted. Security researchers have done much work to prevent attacks with high power algorithms, such as the international effort to develop the current Advanced Encryption Standard (AES). Unfortunately, IoT devices do not typically have the computational resources to implement high-power algorithms such as AES, and must rely on lightweight primitives such as pseudorandom number generators, or PRNGs.This thesis explores the effectiveness, functionality, and use of PRNGs in different applications. First, this thesis investigates the confidentiality of a single-stage residue number system PRNG, which has previously been shown to provide extremely high quality outputs for simulation and digital communication applications when evaluated through traditional techniques like the battery of statistical tests used in the NIST Random Number Generation and DIEHARD test suites or in using Shannon entropy metrics. In contrast, rather than blindly performing statistical analyses on the outputs of the single-stage RNS PRNG, this thesis provides both white box and black box analyses that facilitate reverse engineering of the underlying RNS number generation algorithm to obtain the residues, or equivalently the key, of the RNS algorithm. This thesis develops and demonstrate a conditional entropy analysis that permits extraction of the key given a priori knowledge of state transitions as well as reverse engineering of the RNS PRNG algorithm and parameters (but not the key) in problems where the multiplicative RNS characteristic is too large to obtain a priori state transitions. This thesis then discusses multiple defenses and perturbations for the RNS system that defeat the original attack algorithm, including deliberate noise injection and code hopping. We present a modification to the algorithm that accounts for deliberate noise, but rapidly increases the search space and complexity. Lastly, a comparison of memory requirements and time required for the attacker and defender to maintain these defenses is presented. The next application of PRNGs is in building a translation for binary PRNGs to non-binary uses like card shuffling in a casino. This thesis explores a shuffler algorithm that utilizes RNS in Fisher-Yates shuffles, and that calls for inputs from any PRNG. Entropy is lost through this algorithm by the use of PRNG in lieu of TRNG and by its RNS component: a surjective mapping from a large domain of size $2^J$ to a substantially smaller set of arbitrary size $n$. Previous research on the specific RNS mapping process had developed a lower bound on the Shannon entropy loss from such a mapping, but this bound eliminates the mixed-radix component of the original formulation. This thesis calculates a more precise formula which takes into account the radix, $n$. This formulation is later used to specify the optimal parameters to simulate the shuffler with different test PRNGs. After implementing the shuffler with PRNGs with varying output entropies, the thesis examines the output value frequencies to discuss if utilizing PRNG is a feasible alternative for casinos to the higher-cost TRNG. / Master of Science / Cryptography, or the encrypting of data, has drawn widespread interest for years, initially sparking public concern through headlines and dramatized reenactments of hackers targeting security protocols. Previous cryptographic research commonly focused on developing the quickest, most secure ways to encrypt information on high-power computers. However, as wireless low-power devices such as smart home, security sensors, and learning thermostats gain popularity in ordinary life, interest is rising in protecting information being sent between devices that don't necessarily have the power and capabilities as those in a government facility. Lightweight primitives, the algorithms used to encrypt information between low-power devices, are one solution to this concern, though they are more susceptible to attackers who wish to reverse engineer the encrypting process. The pesudorandom number generator (PRNG) is a type of lightweight primitive that generates numbers that are essentially random even though it is possible to determine the input value, or seed, from the resulting output values. This thesis explores the effectiveness and functionality of PRNGs in different applications. First, this thesis explores a PRNG that has passed many statistical tests to prove its output values are random enough for certain applications. This project analyzes the quality of this PRNG through a new lens: its resistance to reverse engineering attacks. The thesis describes and implements an attack on the PRNG that allows an individual to reverse engineer the initial seed. The thesis then changes perspective from attacker to designer and develop defenses to this attack: by slightly modifying the algorithm, the designer can ensure that the reverse engineering process is so complex, time-consuming, and memory-requiring that implementing such an attack would be impractical for an attacker. The next application of PRNGs is in the casino industry, in which low-power and cost-effective automatic card shufflers for games like poker are becoming popular. This thesis explores a solution for optimal shuffling of a deck of cards.

The main goal of this research is to facilitate end-user and expert developer collaboration in the creation of a web application. This research created a reverse engineering toolset and integrated it with Click (Component-based Lightweight Internet-application Construction Kit), an end-user web development tool. The toolset generates artifacts to facilitate collaboration between end-users and expert web developers when the end-users need to go beyond the limited capabilities of Click. By supporting smooth transition of workflow to expert web developers, we can help them in implementing advanced functionality in end-user developed web applications. The four artifacts generated include a sitemap, text documentation, a task model, and a canonical representation of the user interface. The sitemap is automatically generated to support the workflow of web developers. The text documentation of a web application is generated to document data representation and business logic. A task model, expressed using ConcurTaskTrees notation, covers the whole interaction specified by the end-user. A presentation and dialog model, represented in User Interface Markup Language (UIML), describe the user interface in a declarative language. The task model and UIML representation are created to support development of multi-platform user interfaces from an end-user web application. A formative evaluation of the usability of these models and representations with experienced web developers revealed that these representations were useful and easy to understand. / Master of Science

In education and research, references play a key role. They give credit to prior works, and provide support for reviews, discussions, and arguments. The set of references attached to a publication can help describe that publication, can aid with its categorization and retrieval, can support bibliometric studies, and can guide interested readers and researchers. If suitably analyzed, that set can aid with the analysis of the publication itself, especially regarding all its citing passages. However, extracting and parsing references are difficult problems. One concern is that there are many styles of references, and identifying what style was employed is problematic, especially in heterogeneous collections of theses and dissertations, which cover many fields and disciplines, and where different styles may be used even in the same publication. We address these problems by drawing upon suitable knowledge found in the WWW. In particular, we use appropriate lists (e.g., of names, cities, and other types of entities). We use available information about the many reference styles found, in a type of reverse engineering. We use available references to guide machine learning. In particular, we research a two-stage classifier approach, with multi-class classification with respect to reference styles, and partially solve the problem of parsing surface representations of references. We describe empirical evidence for the effectiveness of our approach and plans for improvement of our method. / Ph. D.

<p dir="ltr">Artificial intelligence (AI) is increasingly integrated into critical systems across various sectors, including public surveillance, autonomous driving, and malware detection. Despite their impressive performance and promise, the security and safety of AI-enabled systems remain significant concerns. Like conventional systems that have software bugs or vulnerabilities, applications leveraging AI are also susceptible to such issues. Malicious behaviors can be intentionally injected into AI models by adversaries, creating a backdoor. These models operate normally with benign inputs but consistently misclassify samples containing an attacker-inserted trigger, known as a <i>backdoor attack</i>.</p><p dir="ltr">However, backdoors can not only be injected by an attacker but may also naturally exist in normally trained models. One can find backdoor triggers in benign models that cause any inputs with the trigger to be misclassified, a phenomenon termed <i>natural backdoors</i>. Regardless of whether they are injected or natural, backdoors can take various forms, which increases the difficulty of identifying such vulnerabilities. This challenge is exacerbated when access to AI models is limited.</p><p dir="ltr">This dissertation introduces an optimization-based technique that reverse-engineers trigger patterns exploited by backdoors, whether injected or natural. It formulates how backdoor triggers modify inputs down to the pixel level to approximate their potential forms. The intended changes in output predictions guide the reverse-engineering process, which involves computing the input gradient or sampling possible perturbations when model access is limited. Although various types of backdoors exist, this dissertation demonstrates that they can be effectively clustered into two categories based on their methods of input manipulation. The development of practical reverse-engineering approaches is based on this fundamental classification, leading to the successful identification of backdoor vulnerabilities in AI models.</p><p dir="ltr">To alleviate such security threats, this dissertation introduces a novel hardening technique that enhances the robustness of models against adversary exploitation. It sheds light on the existence of backdoors, which can often be attributed to the small distance between two classes. Based on this analysis, a class distance hardening method is proposed to proactively enlarge the distance between every pair of classes in a model. This method is effective in eliminating both injected and natural backdoors in a variety of forms.</p><p dir="ltr">This dissertation aims to highlight both existing and newly identified security and safety challenges in AI systems. It introduces novel formulations of backdoor trigger patterns and provides a fundamental understanding of backdoor vulnerabilities, paving the way for the development of safer and more secure AI systems.</p>

A common problem experienced by the software engineering community traditionally has been that of understanding legacy code. A decade ago, legacy code was used to refer to programs written in COBOL, typically for large mainframe systems. However, current software developers predominantly use Object Oriented languages like C++ and Java. The belief prevalent among software developers and object philosophers that comprehending object-oriented software will be relatively easier has turned out to be a myth. Tomorrow's legacy code is being written today, since object oriented programs are even more complex and difficult to comprehend, unless rigorously documented. Reverse Engineering is a methodology that greatly reduces the time, effort and complexity involved in solving the program comprehension problem. This thesis deals with Reverse Engineering complex object oriented software and the experiences with a sample case study. Extensive survey of literature and contemporary research on reverse engineering and program comprehension was undertaken as part of this thesis work. An Energy Information System (EIS) application created by a leading energy service provider and one that is being used extensively in the real world was chosen as a case study. Reverse engineering this industry strength Java application necessitated the definition of a formal process. An intuitive Reverse Engineering Process (REP) was defined and used for the reverse engineering effort. The learning experiences gained from this case study are discussed in this thesis. / Master of Science

A key advantage of Additive Manufacturing (AM) is the opportunity to design and fabricate complex structures that cannot be made via traditional means. However, this potential is significantly constrained by the use of a facet-based geometry representation (e.g., the STL and the AMF file formats); which do not contain any volumetric information and often, designing/slicing/printing complex geometries exceeds the computational power available to the designer and the AM system itself. To enable efficient design and fabrication of complex/multi-material complex structures, several algorithms are presented that represent and process solid models as a set of voxels (three-dimensional pixels). Through this, one is able to efficiently realize parts featuring complex geometries and functionally graded materials. This thesis specifically aims to explore applications in three distinct fields namely, (i) Design for AM, (ii) Design for Manufacturing (DFM) education, and (iii) Reverse engineering from imaging data wherein voxel-based representations have proven to be superior to the traditional AM digital workflow. The advantages demonstrated in this study cannot be easily achieved using traditional AM workflows, and hence this work emphasizes the need for development of new voxel based frameworks and systems to fully utilize the capabilities of AM. / MS / Additive Manufacturing(AM) (also referred to as 3D Printing) is a process by which 3D objects are constructed by successively forming one-part cross-section at a time. Typically, the input file format for most AM systems is in the form of surface representation format (most commonly. STL file format). A STL file is a triangular representation of a 3-dimensional surface geometry where the part surface is broken down logically into a series of small triangles (facets). A key advantage of Additive Manufacturing is the opportunity to design and fabricate complex structures that cannot be made easily via traditional manufacturing techniques. However, this potential is significantly constrained by the use of a facet-based (triangular) geometry representation (e.g., the STL file format described above); which does not contain any volumetric (for e.g. material, texture, color etc.) information. Also, often, designing/slicing/printing complex geometries using these file formats can be computationally expensive. To enable more efficient design and fabrication of complex/multi-material structures, several algorithms are presented that represent and process solid models as a set of voxels (three-dimensional pixels). A voxel represents the smallest representable element of volume. For binary voxel model, a value of ‘1’ means that voxel is ‘on’ and value of 0 means voxel is ‘off’. Through this, one is able to efficiently realize parts featuring complex geometries with multiple materials. This thesis specifically aims to explore applications in three distinct fields namely, (i) Design for AM, (ii) Design for Manufacturing (DFM) education, and (iii) Fabricating models (Reverse engineering) directly from imaging data. In the first part of the thesis, a software tool is developed for automated manufacturability analysis of a part that is to be produced by AM. Through a series of simple computations, the tool provides feedback on infeasible features, amount of support material, optimum orientation and manufacturing time for fabricating the part. The results from this tool were successfully validated using a simple case study and comparison with an existing pre-processing AM software. Next, the above developed software tool is implemented for teaching instruction in a sophomore undergraduate classroom to improve students’ understanding of design constraints in Additive Manufacturing. Assessments are conducted to measure students’ understanding of a variety of topics in manufacturability both before and after the study to measure the effectiveness of this approach. The third and final part of this thesis aims to explore fabrication of models directly from medical imaging data (like CT Scan and MRI). A novel framework is proposed which is validated by fabricating three distinct medical models: a mouse skull, a partial human skull and a horse leg directly from corresponding CT Scan data. The advantages demonstrated in this thesis cannot be easily achieved using traditional AM workflows, and hence this work emphasizes the need for development of new voxel based frameworks and systems to fully utilize the capabilities of AM.

Die vorliegende Arbeit befasst sich mit der Entwicklung und Umsetzung von Methoden und Werkzeugen zur Bereitstellung von Modellen und Randbedingungen für die Konstruktion individueller Langzeit-Knochenimplantate (Konstruktionsvorbereitung). Grundlage dabei ist, dass die Planung aus medizinischer Sicht z.B. durch einen Chirurgen und die Konstruktion unter technischen Aspekten z.B. durch einen Konstrukteur getrennt erfolgt. Hierfür wird ein erarbeitetes Planungskonzept vorgestellt, welches sowohl die geplanten geometrischen Merkmale, als auch weiterführende Metadaten beinhaltet (Randbedingungen). Die Übergabe dieser Planungsdaten an die Konstruktion erfolgt über eine dafür entworfene Formatbeschreibung im Kontext der Schnittstelle zwischen Mediziner und Ingenieur. Weiterführend wird die Notwendigkeit von speziellen Funktionen für die Konstruktion von individuellen Implantaten in der Arbeitsumgebung des Konstrukteurs (z.B. Modelliersystem – CAD) am Beispiel der konturlinienbasierten Modellrekonstruktion erörtert. Die gesamtheitliche Basis bildet eine durchgängig digitale Prozesskette zur Datenaufbereitung, Konstruktion und Fertigung. Die Anwendbarkeit der Methoden und zweier umgesetzter Demonstratoren wurde innerhalb eines interdisziplinär angelegten Projektes am realen Patientenfall bestätigt.