VTechWorks staff will be away for the Thanksgiving holiday beginning at noon on Wednesday, November 27, through Friday, November 29. We will resume normal operations on Monday, December 2. Thank you for your patience.

Browsing by Author "Cody, Tyler Michael"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Closed System Precepts in Systems Engineering for Artificial Intelligence- SE4AI
    Shadab, Niloofar (Virginia Tech, 2024-01-08)
    Intelligent systems ought to be distinguished as a special type of systems that require distinctive engineering processes. While this distinction is informally acknowledged by some, practical systems engineering (SE) methodologies for intelligent systems remain primarily rooted in traditional SE paradigms centered around component aggregation. Initially, this dissertation posits that the traditional approach is grounded in the notion of open systems as the fundamental precept, whereas engineering intelligent systems necessitates an alternative approach founded on the principles of closed systems. This dissertation endeavors to identify potential gaps within the current SE foundations concerning the accommodation of the unique characteristics of intelligent systems, such as continuous learning and sensitivity to environmental changes. Furthermore, it argues for the mitigation of these gaps through the formalization of closed systems precepts. It adopts a systems-theoretic perspective to elucidate the phenomena of closed systems and their intricate interplay with engineering intelligent systems. This research contends that, given the intricate coupling between intelligent systems and their environments, the incorporation of closed systems precepts into SE represents a pivotal pathway to construct engineered intelligence. In pursuit of this objective, this dissertation establishes a formal foundation to delineate closed systems precepts and other fundamental practices. Subsequently, it provides formalism to discern two important categories of closed systems, informationally and functionally closed systems, and their relevance in the domains of engineering and design across diverse levels of system abstraction. Additionally, it explores the practical application of the closed systems precepts through the novel paradigm of core and periphery, followed by its examination within real-world contexts. This dissertation is organizes as follows: Chapter 1 initiates the dissertation by presenting the problem formulation and motivation. It subsequently delves into a thorough literature review and outlines the research's scope and objectives, contributing to the essence of this work. In Chapter 2, a narrative unfolds, elucidating the contributions of the provided papers to the objectives outlined in Chapter 1. This chapter illuminates how each paper aligns with and furthers the overarching goals set forth in the Chapter 1. Chapter 3 serves as a culmination, offering a summary of the accomplishments, acknowledging limitations, and delineating potential avenues for future research within this domain. Paper A is devoted to substantiating the closed notion of intelligence property. In the realm of artificial intelligence (AI), systems are often expected to exert influence upon their environments and, reciprocally, to be influenced by their surroundings. Consequently, a profound interdependence exists between the system and its environment, transcending the confines of conventional input-output relations. In this regard, Paper A postulates that the engineering of intelligent systems mandates an approach that elevates closed systems as foundational precepts for characterizing intelligence as a property contingent upon the system's relationship with its context. The ensuing discussion will juxtapose the viewpoints of open and closed systems, illustrating the limitations of the open system perspective in representing intelligence as a relational property. In response, this paper will advocate for the adoption of the closed system view to establish intelligence as an inherent relational property arising from the system's dynamic interactions with its environment. Paper B is dedicated to the formalization of the closed systems paradigm within SE. In this paper, formalism is proffered for the closed systems precepts, drawing upon systems theory, cybernetics, and information theory. A comprehensive comparison of two closure types, informational and functional closure, within closed systems is presented, underpinned by a common systems-theoretic formal framework. This dissertation contends that by grounding these initiatives in the core and periphery concept, we can facilitate the design and engineering of intelligent systems across multiple levels of abstraction. These levels may span a spectrum from informational closure to a synthesis of informational and functional openness. It posits that this approach represents a versatile, method-agnostic solution to some of the principal challenges encountered when engineering multiple tiers of intelligence for complex systems. Paper C delves into the rise of the concept of core-periphery from some cybernetics principles, such as variety and "The Law of Requisite Variety" and provides a formalism that is a derivation of the mentioned principles in Cybernetics. Later, it elaborates on the practical implications of such concepts in intelligent systems from biological systems and entails an engagement with a CNN model to explore the core and periphery concept within AI-enabled systems. Paper D proposes the practical implementation of the closed systems doctrine in SE, offering frameworks that rigorously define the boundaries between closed systems and their environment. These frameworks are meticulously designed to account for stakeholder requirements and the inherent design constraints of the system. This paper illustrates practical applications of informational and functional closure within SE processes, leveraging a hypothetical example for elucidation. It focuses on two aspects of engineering intelligence, scope and scale to provide a platform for the utilization of closed systems precepts.
  • Thumbnail Image
    RESONANT: Reinforcement Learning Based Moving Target Defense for Detecting Credit Card Fraud
    Abdel Messih, George Ibrahim (Virginia Tech, 2023-12-20)
    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).