Doctor of Philosophy, The Ohio State University, 2022, Computer Science and Engineering

With the continuous increase of time series data, more and more research is focused on using these data to improve people's lives. On the one hand, the Markov Decision Process (MDP) is used widely in decision-making. An agent can decide the best action based on its current state. When the agent is applied to time series data, the model will help people make more informed decisions. However, state identification, which is very important in obtaining an optimal decision, has received less attention. On the other hand, with the development of deep learning, identifying the category of a time series has become more and more precise. As a result, the recognition of complex time series sequences has become the hub of public attention. In this dissertation, we focus on developing an automatic state selection using MDP and investigate the application of deep learning in recognizing time series data. We propose a method that combines decision-tree modeling and MDP to permit automatic state identification in a way that offers desirable trade-offs between simplicity and Markovian behavior. We first create a simplified definition of the host state, which becomes the response measure in our decision-tree model. Then, we fit the model in a way that weighs accuracy and interpretability. The leaves of the resulting decision-tree model become the system states. This follows, intuitively, because these are the groupings needed to predict (approximately) the system evolution. Then, we generate and apply an MDP control policy. Our motivating example is cyber vulnerability maintenance. Using the proposed methods, we predict that a Midwest university could save more than four million dollars compared to the current policy. Prechtl's general movements assessment (GMA) allows visual recognition of movement patterns in infants that, when abnormal (cramped synchronized, or CS), have very high specificity in predicting later neuromotor disorders. However, training req (open full item for complete abstract)

Committee: Rajiv Ramnath (Advisor); Ping Zhang (Committee Member); Theodore Allen (Advisor) Subjects: Artificial Intelligence; Bioinformatics; Business Costs; Computer Engineering; Computer Science

Master of Science, The Ohio State University, 2021, Computer Science and Engineering

Digital contact tracing offers significant promise to help reduce the spread of SARS-CoV-2 and other viruses. Google and Apple joined together to create the Google/Apple Exposure Notification (GAEN) framework to determine encounters with anonymous users later diagnosed COVID-19 positive. However, as GAEN lacks geospatial awareness, it is susceptible to geographically distributed replay attacks. While the replay attack is generally known, we contribute a new proof-of-concept for an easily deployed, anonymous, low-cost, crowd-sourced replay attack network by malicious actors (or far away nation-state attackers) who utilize malicious (or innocent) users' smartphones to capture and replay GAEN advertisements that drastically increase false-positive rates even in areas that otherwise exhibit low positivity rates. In response to this powerful and feasible replay attack, we introduce GAEN+, a solution that enhances GAEN with geospatial awareness while maintaining user privacy, and demonstrate its ability to effectively prevent distributed replay attacks with negligible overhead compared with the original GAEN framework.

Committee: Anish Arora (Advisor); Zhiqiang Lin (Committee Member) Subjects: Computer Science

Master of Science, The Ohio State University, 2017, Industrial and Systems Engineering

Cyber vulnerability can be exploited by cyber-attackers to achieve valuable information, alter or destroy a cyber-target. Finding a way to generate appropriate cyber vulnerability maintenance policies (a combination of maintenance actions) is crucial for cyber security administrators. The purpose of this thesis is to apply a data-driven Markov decision processes model to generate cyber vulnerability policies that minimize administrative costs, including maintenance action cost and incident risk cost, in the long term. Optimal policies aim if not to eliminate then at least to reduce the incident risk to an acceptable level. By exploiting the real-world data of Nessus scan reports and incident reports from the OSU, a host-based dataset is built to analyze the characteristics of hosts and develop host-based policies. After solving the MDP model, the optimal policies and related costs are presented in comparison with existing policy. The results show that, for hosts in management groups, the incident risk and action costs are significantly lower than for hosts with administrative privilege, and more advanced actions can be taken to protect the hosts from cyber-attacks as the result of the discounted action costs. The consequences of a successful intrusion into a critical server are more serious than for a normal host, therefore, more powerful actions are required for critical servers. For the remainder of hosts, applying only auto patching is recommended for most situations, especially for non-general-purpose hosts such as printers and routers.

Committee: Theodore Allen (Advisor); Cathy Xia (Committee Member) Subjects: Operations Research

Doctor of Philosophy, The Ohio State University, 2017, Industrial and Systems Engineering

This research explores deterministic and stochastic policies to help organizations make data-driven optimal decisions. The two major application areas identified in this research are manufacturing and cyber security. In a recent report published by McKinsey Analytics, the manufacturing industry uses only 20%-30% of the potential of data analytics. This suggests that there are still plenty of opportunities to use analytics in manufacturing processes. In the first part of my research, I formulate an Integer Programming model for the “stamping” process in automotive manufacturing. I develop a production scheduling method for automotive stamping to maintain optimal inventory positions. In stamping, different types of parts are scheduled for processing in the press, which requires different die-sets to be mounted on the press. This has all the elements of conventional scheduling problems with tardiness objectives and setup costs. Yet, it also has capacity constraints and part production constraints. We show that these constraints make solution with branch and bound difficult for problem sizes of interest. In this research, I use the structure of the scheduling problem and implemented heuristic methods like Genetic Algorithm alongside Earliest Due-date (EDD) rules to prioritize production of parts with low inventory as well as minimize the number of die-set changeovers. I call this new method Genetic Algorithm with Generalized Earliest Due-date (GAGEDD). I illustrate the computational advantages compared with alternatives and show its benefits using data from a real life automotive stamping press scheduling problem to build a decision support tool for the schedulers. The second part of this research is motivated towards improving cyber vulnerability maintenance policies under uncertainty. A conservative estimate by McAfee in 2014 puts annual cost of cybercrime at US$375B. This is an important contemporary issue where role of data analytics and optimization have a lot (open full item for complete abstract)

Committee: Theodore T. Allen PhD (Advisor); Cathy H. Xia PhD (Committee Member); Gagan Agrawal PhD (Committee Member) Subjects: Industrial Engineering; Operations Research

Master of Science in Computer Engineering (MSCE), Wright State University, 2015, Computer Engineering

Designing secure cyber-physical systems (CPS) is fundamentally important. An indispensable step towards this end is to perform vulnerability assessment. This thesis discusses the design and implementation of a mission-aware CPS vulnerability assessment framework. The framework intends to accomplish three objectives including i) mapping CPS mission into infrastructural components, ii) evaluating global impact of each vulnerability, and iii) achieving verifiable results and high flexibility. In order to accomplish these objectives, a model-based analysis strategy is employed. Specifically, a CPS simulator is used to model dynamic behaviors of CPS components under different missions; the framework facilitates a bottom-up approach to traverse a holistic model of a CPS that aims at profiling relationships among all CPS components. In order to analyze the derived models, we have leveraged formal methods, including program symbolic execution, logic programming, and linear optimization. The framework first successfully identifies mission-critical components, then discovers all attack paths from system access points to mission-critical components, and finally recommends the optimized mitigation plan.

Committee: Junjie Zhang Ph.D. (Advisor); Adam Robert Bryant Ph.D. (Committee Member); Michelle Andreen Cheatham Ph.D. (Committee Member) Subjects: Computer Engineering; Computer Science