Artificial Intelligence-Enabled Information Privacy Analytics and Risk Assessment

Abstract

In the contemporary digital ecosystem, the proliferation of personal data accessibility has markedly amplified concerns surrounding information privacy. Despite the critical nature of these concerns, extant measures (e.g., privacy policies, privacy settings) for safeguarding information privacy have demonstrated substantial deficiencies. Consequently, there exists an imperative need for the enhancement of information privacy analytics and risk assessment methodologies. Such advancements are essential to augment individuals' understanding and vigilance regarding their personal information exposure and the attendant privacy risks.This dissertation addresses this exigent issue through the development and implementation of Artificial Intelligence (AI)-enabled analytical tools aimed at dissecting privacy data. The ultimate goal is to employ the insights derived from these analyses to bolster efficient, proactive, and informed privacy defenses, with a particular focus on populations deemed at risk. Structured into four distinct essays, this scholarly work unfolds as follows: The inaugural essay delineates the creation of a Deep Learning (DL)-based text classification system tasked with segmenting and categorizing privacy policy content. This system serves to evaluate the dynamic nature of privacy policies in light of evolving privacy regulations. Subsequently, the second essay introduces a DL-based Entity Resolution (ER) system engineered to correlate Personally Identifiable Information (PII) surfaced on Dark Web platforms with corresponding data on Surface Web platforms, thereby facilitating a holistic assessment of individuals' information privacy risk profiles. The third essay extends the theoretical framework of the Protection Motivation Theory (PMT) to underpin the design of an exposed PII search mechanism aimed at succinctly communicating privacy threats and countermeasures to foster privacy-protective behaviors among senior citizens. The fourth essay proposes the implementation of a Large Language Model (LLM)-based User Identity Linkage (UIL) system designed to interlink social media accounts across multiple platforms to evaluate social media surveillance risks comprehensively. By integrating AI methodologies with the discipline of information privacy, this dissertation endeavors to chart a path toward the development of cutting-edge information privacy protection mechanisms. Adhering to the principles of the design science paradigm and integrating insights gleaned from behavioral research, the models, frameworks, design principles, and theories elucidated within this dissertation aim to make substantive contributions to the field of Information Systems (IS) information privacy research.